← All changes | otto/otto_pixel.php +1200 -319 2.6.15 → 2.7.2 View file →

		@@ -16,18 +16,67 @@
16	16	# uses the simple html dom library
17	17	use simplehtmldom\HtmlDocument;
18	18
19	19	# include the otto class file
20		-require_once plugin_dir_path( __FILE__ ) . '/vendor/autoload.php';
	20	+#
	21	+# The vendor tree is generated content, unlike the plugin files beside it: an
	22	+# update that copies the plugin directory piece by piece can briefly leave this
	23	+# file beside an otto/vendor that is missing, half-written or holding a stale
	24	+# classmap. A bare require on a missing autoloader is a fatal during plugin
	25	+# bootstrap — the whole site, not just OTTO. Load it only when it is actually
	26	+# readable, and catch Throwable while doing so: the autoloader is a trampoline
	27	+# of bare requires into otto/vendor/composer/, so a tree that is present but
	28	+# still mid-copy (or a truncated autoload.php, which parses only to a
	29	+# ParseError) would otherwise fatal here too. Either way the parser class
	30	+# stays unloaded, and the DOM-parser guard in metasync_start_otto() degrades
	31	+# the request to OTTO-less rendering instead of fataling.
	32	+$metasync_otto_vendor_autoload = plugin_dir_path( __FILE__ ) . '/vendor/autoload.php';
	33	+if ( is_readable( $metasync_otto_vendor_autoload ) ) {
	34	+ try {
	35	+ require_once $metasync_otto_vendor_autoload;
	36	+ } catch ( \Throwable $e ) {
	37	+ # Deliberately swallowed: any exception raised while loading generated
	38	+ # autoloader code means this request runs without the DOM parser. The
	39	+ # guard below turns that into one OTTO-less request rather than a
	40	+ # bootstrap fatal on the whole site.
	41	+ }
	42	+}
21	43	require_once plugin_dir_path( __FILE__ ) . '/Otto_html_class.php';
22	44	require_once plugin_dir_path( __FILE__ ) . '/Otto_pixel_class.php';
23	45	require_once plugin_dir_path( __FILE__ ) . '/metasync-otto-seo-functions.php';
24	46	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-transient-cache.php';
	47	+require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-job-status.php';
	48	+require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-webhook-audit.php';
25	49	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-render-strategy.php';
26	50	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-config.php';
27	51	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-bot-detector.php';
28	52	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-bot-statistics-database.php';
29	53
	54	+/**
	55	+ * Is $method actually available on $class in this process right now?
	56	+ *
	57	+ * A partially updated install can leave a newer copy of one plugin file beside
	58	+ * an older copy of another — stale opcache bytecode for a single file is enough.
	59	+ * class_exists() is satisfied by the older copy, and calling a method it does
	60	+ * not declare is a fatal. Everything in this file runs on the front end, so that
	61	+ * fatal is a white screen on every page view rather than a degraded feature.
	62	+ *
	63	+ * Defined here rather than pulled from a shared utility class on purpose: a
	64	+ * helper loaded from another file could itself be the stale one.
	65	+ *
	66	+ * $class and $method are parameters rather than literals at the call site so the
	67	+ * check survives static analysis, which would otherwise narrow a literal
	68	+ * method_exists() on a known class to a constant true — the skew this guards
	69	+ * against exists only at runtime.
	70	+ *
	71	+ * @param string $class Class about to be called.
	72	+ * @param string $method Method about to be called on it.
	73	+ * @return bool
	74	+ */
	75	+function metasync_otto_class_provides($class, $method) {
	76	+ return class_exists($class) && method_exists($class, $method);
	77	+}
	78	+
30	79	# OPTIMIZED: get the metasync options (cached in static class)
31	80	$metasync_options = Metasync_Otto_Config::get_options();
32	81
33	82	# OTTO SSR is always enabled by default
		@@ -51,9 +100,9 @@
51	100	# Use endpoint manager to get the correct domain
52	101	$otto_domain = 'sa.searchatlas.com'; # default
53	102	if (class_exists('Metasync_Endpoint_Manager')) {
54	103	$otto_api_domain = Metasync_Endpoint_Manager::get_endpoint('OTTO_API_DOMAIN');
55		- $parsed = parse_url($otto_api_domain);
	104	+ $parsed = wp_parse_url($otto_api_domain);
56	105	if (!empty($parsed['host'])) {
57	106	$otto_domain = $parsed['host'];
58	107	}
59	108	}
		@@ -65,12 +114,54 @@
65	114	# OPTIMIZED: use cached uuid
66	115	$otto_tag = '<meta name="otto" content="uuid='.esc_attr(Metasync_Otto_Config::get_otto_uuid()).'; type=wordpress; enabled='.esc_attr($string_enabled).'; version='.esc_attr($plugin_version).'">';
67	116
68	117	# out the otto tag
	118	+ // phpcs:ignore WordPress.Security.EscapeOutput.OutputNotEscaped -- meta tag assembled on the line above with esc_attr() on every interpolated value
69	119	echo $otto_tag;
70	120	}, 1); # Priority 1 to output early in head
71	121
72	122	/**
	123	+ * Scheduling seam for the OTTO pipeline.
	124	+ *
	125	+ * wp_schedule_single_event() returns false on failure and the old pipeline
	126	+ * ignored it, so a broken cron table acknowledged URLs that were never
	127	+ * queued. Every schedule in this pipeline goes through this seam so the
	128	+ * return value is honored (and tests can observe scheduling).
	129	+ */
	130	+function metasync_otto_schedule_single_event($timestamp, $hook, $args = array()) {
	131	+ return wp_schedule_single_event($timestamp, $hook, $args) === true;
	132	+}
	133	+
	134	+/**
	135	+ * Dedupe seam: does this URL already have live per-URL work in flight?
	136	+ *
	137	+ * Combines the durable status store with the raw cron table so a re-delivered
	138	+ * webhook cannot stack a second job behind a live one.
	139	+ */
	140	+function metasync_otto_job_already_scheduled($route) {
	141	+ if (class_exists('Metasync_Otto_Job_Status', false)
	142	+ && Metasync_Otto_Job_Status::has_active($route)) {
	143	+ return true;
	144	+ }
	145	+ return wp_next_scheduled('metasync_process_otto_crawl_url_job', array($route)) !== false;
	146	+}
	147	+
	148	+/**
	149	+ * Cron-table dedupe seam: does a per-URL cron event already exist for this URL?
	150	+ *
	151	+ * The pending-queue drainer must dedupe against the scheduler table only.
	152	+ * Every URL it takes off the durable queue carries a fresh webhook
	153	+ * accepted/batch_overflow ledger row; that row exists so a re-delivered
	154	+ * webhook cannot stack a second job behind live work, and consulting it
	155	+ * from the drainer — after take_pending() has already removed the URL
	156	+ * from the queue — silently dropped every overflow URL instead of
	157	+ * draining it. Only a real, still-pending cron event may veto here.
	158	+ */
	159	+function metasync_otto_job_event_pending($route) {
	160	+ return wp_next_scheduled('metasync_process_otto_crawl_url_job', array($route)) !== false;
	161	+}
	162	+
	163	+/**
73	164	* Start end point to handle requests on page updates
74	165	* The Otto Crawler will call this end point once a page is updated
75	166	**/
76	167
		@@ -76,8 +167,13 @@
76	167
77	168	# function to register the route
78	169	function metasync_otto_crawl_notify($request){
79	170
	171	+ # receipt time is captured before validation and scheduling, so the audit
	172	+ # record describes when WordPress received the delivery, not when its work
	173	+ # happened to finish.
	174	+ $received_at = time();
	175	+
80	176	# get request data params
81	177	$data = $request->get_json_params();
82	178
83	179	# fields
		@@ -97,36 +193,48 @@
97	193	}
98	194
99	195	}
100	196
101		- # load otto pixel
102		- $otto_pixel = new Metasync_otto_pixel(false);
	197	+ # The audit trail is deliberately after payload validation: invalid requests
	198	+ # receive the existing 400 response and are not recorded as deliveries.
	199	+ # It is also before any scheduling so a valid delivery is captured even if
	200	+ # later work is deferred or rejected. The failure guard is intentionally
	201	+ # broader than Exception: a broken option adapter or stale plugin copy must
	202	+ # never turn an otherwise acknowledged webhook into a 500 response.
	203	+ try {
	204	+ if (metasync_otto_class_provides('Metasync_Otto_Webhook_Audit', 'record')) {
	205	+ Metasync_Otto_Webhook_Audit::record(
	206	+ $received_at,
	207	+ $data['domain'],
	208	+ $data['urls']
	209	+ );
	210	+ }
	211	+ } catch (Throwable $metasync_otto_audit_failure) {
	212	+ error_log(
	213	+ 'MetaSync OTTO: failed to record crawl-notify webhook audit entry: '
	214	+ . $metasync_otto_audit_failure->getMessage()
	215	+ ); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- audit failure must not break webhook acknowledgement
	216	+ }
103	217
104		- # save the otto data first (cheap local DB write — kept synchronous)
105		- $otto_pixel->save_crawl_data($data);
106		-
107	218	# Defer the expensive per-URL work (OTTO API fetch, post meta sync,
108	219	# per-URL cache purge) to background jobs. The webhook caller enforces a
109		- # strict response-time budget; doing this work inline previously blocked
110		- # the response for up to 30s × N URLs and timed the caller out.
111		- # Each URL gets its own scheduled event so failures are isolated.
	220	+ # strict response-time budget, so the response acknowledges SCHEDULING,
	221	+ # never execution. The per-URL status store records what actually
	222	+ # happened to each URL once the jobs run.
112	223	#
113		- # WP-299: Cap the number of per-URL cron jobs scheduled per webhook batch.
114		- # Large crawl batches (100+ URLs) previously created hundreds of cron events
115		- # that overwhelmed WP-Cron on shared/managed hosts like WP Engine. Excess
116		- # URLs beyond the cap are silently skipped — OTTO will re-crawl them on the
117		- # next webhook cycle.
	224	+ # Cap the number of per-URL cron jobs scheduled per webhook batch so a
	225	+ # large crawl cannot overwhelm WP-Cron on shared hosts. URLs past the cap
	226	+ # are queued durably and worked off by a self-rescheduling drainer job —
	227	+ # never silently dropped.
118	228	$max_jobs_per_batch = defined('METASYNC_MAX_JOBS_PER_BATCH') ? METASYNC_MAX_JOBS_PER_BATCH : 25;
119	229	$now = time();
120	230	$routes_to_process = array();
121		- $scheduled_count = 0;
	231	+ $accepted = 0;
	232	+ $deferred = 0;
	233	+ $rejected = 0;
	234	+ $scheduled_this_batch = 0;
122	235
123	236	foreach($data['urls'] AS $key => $url){
124		- # Enforce per-batch cap to prevent cron overload
125		- if ($scheduled_count >= $max_jobs_per_batch) {
126		- break;
127		- }
128		-
129	237	# prepare the route
130	238	$route = $data['domain'] . $url;
131	239
132	240	# validate the route
		@@ -145,39 +253,71 @@
145	253	if (metasync_is_otto_url_excluded($route)) {
146	254	continue;
147	255	}
148	256
	257	+ # Every valid URL belongs to the batched cache-purge set even when its
	258	+ # per-URL meta-sync job is deduped or deferred: the crawler hit all of
	259	+ # them, so the caches for all of them must be refreshed.
	260	+ $routes_to_process[] = $route;
	261	+
	262	+ # Duplicate delivery: an active job already owns this URL. Idempotent
	263	+ # no-op — not re-scheduled, not queued, not counted as new work.
	264	+ if (metasync_otto_job_already_scheduled($route)) {
	265	+ continue;
	266	+ }
	267	+
	268	+ if ($scheduled_this_batch >= $max_jobs_per_batch) {
	269	+ # Batch overflow: queue durably for the drainer instead of dropping.
	270	+ if (Metasync_Otto_Job_Status::enqueue_pending($route)) {
	271	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'queue', 0, 'batch_overflow');
	272	+ $deferred++;
	273	+ } else {
	274	+ $rejected++;
	275	+ }
	276	+ continue;
	277	+ }
	278	+
149	279	# Queue the per-URL processing for background execution. The handler
150	280	# (metasync_handle_otto_crawl_url_job) performs transient warming,
151	281	# SEO meta sync, and per-URL host cache purge.
152	282	# Offset each event by $key seconds to avoid wp_schedule_single_event()
153	283	# silently dropping duplicates when timestamp + hook + args collide.
154		- wp_schedule_single_event($now + $key, 'metasync_process_otto_crawl_url_job', array($route));
155		-
156		- $routes_to_process[] = $route;
157		- $scheduled_count++;
	284	+ # A false return means cron refused the event — fall back to the
	285	+ # durable queue rather than acknowledging a URL nobody will process.
	286	+ if (metasync_otto_schedule_single_event($now + $key, 'metasync_process_otto_crawl_url_job', array($route))) {
	287	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'queue', 0, '');
	288	+ $accepted++;
	289	+ $scheduled_this_batch++;
	290	+ } elseif (Metasync_Otto_Job_Status::enqueue_pending($route)) {
	291	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'queue', 0, 'cron_refused');
	292	+ $deferred++;
	293	+ } else {
	294	+ $rejected++;
	295	+ }
158	296	}
159	297
160		- # Schedule a single batch job for cache warming and edge CDN purge.
161		- # These operations benefit from batching: warm_urls() uses concurrent
162		- # fire-and-forget requests, and edge providers (Cloudflare, Fastly)
163		- # support multi-URL purge in a single API call.
	298	+ # Schedule a single batch job for the edge CDN purge.
	299	+ # The batch waits for per-URL stragglers (retries) before purging, so a
	300	+ # retried URL's fresh content is what reaches the edge. Multi-URL purge
	301	+ # in a single API call keeps edge-provider request counts low.
164	302	if (!empty($routes_to_process)) {
165	303	$batch_time = $now + count($data['urls']) + 5; // run after all per-URL jobs
166		- wp_schedule_single_event($batch_time, 'metasync_process_otto_batch_cache_job', array($routes_to_process));
	304	+ metasync_otto_schedule_single_event($batch_time, 'metasync_process_otto_batch_cache_job', array($routes_to_process));
167	305	}
168	306
169		- # Track OTTO optimization event in GA4 (local, non-blocking)
170		- try {
171		- Metasync_GA4::get_instance()->track_otto_optimization($data);
172		- } catch (Exception $e) {
173		- // Analytics tracking failed, continue
	307	+ # Work off any overflow (or cron-refused URLs) from the durable queue.
	308	+ if ($deferred > 0) {
	309	+ metasync_otto_schedule_single_event($now + 120, 'metasync_otto_pending_drainer', array());
174	310	}
175	311
176		- # Return 200 immediately so the webhook caller does not time out.
	312	+ # Return 200 immediately so the webhook caller does not time out, with an
	313	+ # honest accounting of what happened to the batch.
177	314	return new WP_REST_Response(array(
178	315	'success' => true,
179	316	'message' => 'OTTO crawl notification received',
	317	+ 'accepted' => $accepted,
	318	+ 'deferred' => $deferred,
	319	+ 'rejected' => $rejected,
180	320	), 200);
181	321	}
182	322
183	323	/**
		@@ -182,151 +322,276 @@
182	322
183	323	/**
184	324	* Background handler for a single OTTO crawl-notify URL.
185	325	*
186		- * Runs the per-URL sequence that was previously inline in metasync_otto_crawl_notify():
187		- * warm the OTTO transient cache, sync SEO post meta, and clear the per-URL host cache.
	326	+ * Pipeline per URL:
	327	+ * 1. Warm the OTTO transient cache — this is the ONE live API call the job
	328	+ * makes; the warmed suggestions are handed to the meta sync so the
	329	+ * endpoint is never fetched twice per URL.
	330	+ * 2. Sync SEO post meta (metasync_process_otto_seo_data) and branch on its
	331	+ * outcome: only SUCCESS / NO_CHANGE reach the purge step.
	332	+ * 3. Clear the per-URL host cache and re-warm it.
188	333	*
189		- * Cache warming and edge CDN purge are handled separately by
190		- * metasync_handle_otto_batch_cache_job() in a single batched event, because
191		- * those operations benefit from multi-URL batching (fewer API calls).
	334	+ * Retryable outcomes (API timeout / 5xx / 429 / lock contention / CPU deferral)
	335	+ * retry with exponential backoff up to METASYNC_OTTO_JOB_MAX_RETRIES; the
	336	+ * exhaustion event is recorded via metasync_record_failed_action() for
	337	+ * Site Health. Permanent outcomes (not connected, excluded/404 URL, API auth
	338	+ * or input error) never retry and never purge.
192	339	*
193		- * On failure, retries up to METASYNC_OTTO_JOB_MAX_RETRIES times with 60s backoff.
194		- * After all retries are exhausted the failure is recorded via
195		- * metasync_record_failed_action() for surfacing in Site Health.
196		- *
197	340	* @param string $route Fully-qualified URL to process.
198	341	* @param int $retry_count Current retry attempt (0 = first run).
199	342	*/
200		-function metasync_handle_otto_crawl_url_job($route, $retry_count = 0) {
	343	+function metasync_handle_otto_crawl_url_job($route = '', $retry_count = 0) {
	344	+ // WP-Cron events persist independently of plugin code. A stale or malformed
	345	+ // event may therefore invoke this callback without its required route.
	346	+ // Docblock type isn't enforced at runtime; a stale cron record can pass a non-string.
	347	+ // @phpstan-ignore-next-line function.alreadyNarrowedType
	348	+ if (!is_string($route) || $route === '') {
	349	+ error_log('MetaSync OTTO: skipping crawl-url job without a valid route.'); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
	350	+ return;
	351	+ }
	352	+
	353	+ $max_retries = defined('METASYNC_OTTO_JOB_MAX_RETRIES') ? METASYNC_OTTO_JOB_MAX_RETRIES : 3;
	354	+
	355	+ # Shared retry/terminal bookkeeping so every failure path stays consistent.
	356	+ $schedule_retry = function ($reason) use ($route, $retry_count, $max_retries) {
	357	+ if ($retry_count < $max_retries) {
	358	+ # Exponential backoff (60s, 120s, 240s) plus jitter so a batch of
	359	+ # failures cannot stampede the API on the same tick.
	360	+ $delay = 60 * pow(2, $retry_count) + wp_rand(0, 15);
	361	+ $scheduled = metasync_otto_schedule_single_event(time() + $delay, 'metasync_process_otto_crawl_url_job', array($route, $retry_count + 1));
	362	+ if (!$scheduled && Metasync_Otto_Job_Status::enqueue_pending($route)) {
	363	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'queue', $retry_count + 1, $reason . '/cron_refused');
	364	+ } else {
	365	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_RETRYING, 'job', $retry_count + 1, $reason);
	366	+ }
	367	+ error_log('MetaSync OTTO: retrying crawl-url job for ' . $route . ' (attempt ' . ($retry_count + 1) . '/' . $max_retries . ') reason=' . $reason . ' in ' . $delay . 's'); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
	368	+ } else {
	369	+ metasync_record_failed_action('metasync_process_otto_crawl_url_job');
	370	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_FAILED, 'job', $retry_count, 'exhausted:' . $reason);
	371	+ error_log('MetaSync OTTO: background crawl-url job permanently failed for ' . $route . ' after ' . $max_retries . ' retries, last reason=' . $reason); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
	372	+ }
	373	+ };
	374	+
201	375	try {
202		- # Step 1: Warm OTTO transient cache (fetch fresh suggestions from OTTO API into WP transient)
	376	+ # Not connected: retrying cannot help, and there is nothing to purge.
203	377	$otto_uuid = Metasync_Otto_Config::get_otto_uuid();
204		- if (!empty($otto_uuid)) {
205		- $transient_cache = new Metasync_Otto_Transient_Cache($otto_uuid);
206		- $transient_cache->warm_cache($route);
	378	+ # A corrupted option row can hold a non-string UUID; the docblock
	379	+ # contract alone does not enforce that at runtime.
	380	+ # @phpstan-ignore-next-line function.alreadyNarrowedType
	381	+ if (!is_string($otto_uuid) || $otto_uuid === '') {
	382	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_FAILED, 'job', $retry_count, 'not_connected');
	383	+ return;
207	384	}
208	385
	386	+ # Step 1: Warm OTTO transient cache (fetch fresh suggestions from OTTO
	387	+ # API into WP transient — the rate limiter, circuit breaker, and
	388	+ # request lock all live inside this call).
	389	+ $transient_cache = new Metasync_Otto_Transient_Cache($otto_uuid);
	390	+ $warmed = $transient_cache->warm_cache($route);
	391	+ if ($warmed === false) {
	392	+ # A 4xx refusal (bad key, revoked project) is permanent: retrying
	393	+ # cannot change the answer, so fail fast instead of burning the
	394	+ # attempt budget on warm calls that will keep being rejected.
	395	+ if ($transient_cache->last_failure_is_permanent()) {
	396	+ metasync_record_failed_action('metasync_process_otto_crawl_url_job');
	397	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_FAILED, 'job', $retry_count, 'permanent:warm_rejected');
	398	+ error_log('MetaSync OTTO: crawl-url job failed permanently for ' . $route . ' (warm request rejected by the API)'); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
	399	+ return;
	400	+ }
	401	+ # Transient API failure (timeout / 5xx / breaker open). No meta
	402	+ # write, no purge — straight to the retry path.
	403	+ $schedule_retry('warm_failed');
	404	+ return;
	405	+ }
	406	+
209	407	# Step 2: Write SEO post meta synchronously BEFORE the cache purge.
210		- # This ensures the DB is fully up-to-date when Kinsta (or any host) re-populates
211		- # the cache on the very next request — eliminating the race condition where the
212		- # 1-second scheduled job hadn't run yet and stale meta got cached.
213		- # allow_defer=false: do NOT reschedule a new metasync_process_seo_job cron
214		- # event from this synchronous path — the crawl_url_job retry mechanism already
215		- # handles failures, and rescheduling here caused an unbounded cron pile-up (WP-299).
216		- metasync_process_otto_seo_data($route, false);
	408	+ # This ensures the DB is fully up-to-date when the host re-populates
	409	+ # the cache on the very next request. The warmed suggestions are
	410	+ # passed through so the sync makes NO second API call.
	411	+ # allow_defer=false: the crawl_url_job retry mechanism owns failure
	412	+ # handling here; letting the sync self-reschedule caused an
	413	+ # unbounded cron pile-up.
	414	+ # warm_cache() may hand back a non-array truthy on contract drift;
	415	+ # the sync treats that as "no prefetched suggestions".
	416	+ # @phpstan-ignore-next-line function.alreadyNarrowedType
	417	+ $prefetched = is_array($warmed) ? $warmed : null;
	418	+ $outcome = metasync_process_otto_seo_data($route, false, 0, $prefetched);
217	419
218		- # Step 3: Clear the per-URL cache entry
	420	+ if ($outcome === Metasync_Otto_Job_Status::OUTCOME_RETRYABLE
	421	+ || $outcome === Metasync_Otto_Job_Status::OUTCOME_DEFERRED) {
	422	+ $schedule_retry('outcome_' . $outcome);
	423	+ return;
	424	+ }
	425	+
	426	+ if ($outcome !== Metasync_Otto_Job_Status::OUTCOME_SUCCESS
	427	+ && $outcome !== Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE) {
	428	+ # Permanent failure: no retry, no purge — caches must keep serving
	429	+ # the last good content.
	430	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_FAILED, 'meta', $retry_count, 'permanent:' . $outcome);
	431	+ return;
	432	+ }
	433	+
	434	+ # Step 3: meta is current (changed or confirmed unchanged) — clear the
	435	+ # per-URL host cache entry and re-warm it so OTTO-modified output is
	436	+ # what gets stored.
219	437	$otto_pixel = new Metasync_otto_pixel(false);
220	438	$otto_pixel->refresh_cache($route);
221	439	Metasync_Cache_Purge::purge_single_url($route);
	440	+ Metasync_Cache_Purge::warm_urls(array($route));
222	441
223		- } catch (Exception $e) {
224		- $max_retries = defined('METASYNC_OTTO_JOB_MAX_RETRIES') ? METASYNC_OTTO_JOB_MAX_RETRIES : 3;
	442	+ # A URL that completed on a retry missed its batch's edge purge (the
	443	+ # batch only waits so long). Late completers purge the edge themselves.
	444	+ if ($retry_count > 0) {
	445	+ Metasync_Edge_Cache_Purge::purge(array($route));
	446	+ }
225	447
226		- if ($retry_count < $max_retries) {
227		- # Schedule a retry with exponential backoff: 60s, 120s, 240s
228		- $delay = 60 * pow(2, $retry_count);
229		- wp_schedule_single_event(time() + $delay, 'metasync_process_otto_crawl_url_job', array($route, $retry_count + 1));
230		- error_log('MetaSync OTTO: retrying crawl-url job for ' . $route . ' (attempt ' . ($retry_count + 1) . '/' . $max_retries . ') in ' . $delay . 's: ' . $e->getMessage());
231		- } else {
232		- metasync_record_failed_action('metasync_process_otto_crawl_url_job');
233		- error_log('MetaSync OTTO: background crawl-url job permanently failed for ' . $route . ' after ' . $max_retries . ' retries: ' . $e->getMessage() . ' in ' . $e->getFile() . ':' . $e->getLine());
	448	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_COMPLETED, 'job', $retry_count, $outcome);
	449	+
	450	+ } catch (\Throwable $e) {
	451	+ # Cron must survive both Exception and PHP 7+ Error failures. Always use
	452	+ # the same bounded retry path, and emit structured context without the URL
	453	+ # payload or credentials. The retry is booked BEFORE the log attempt so a
	454	+ # failing logger can never swallow the retry; logging is best-effort.
	455	+ $schedule_retry('throwable:' . get_class($e));
	456	+ if (class_exists('Metasync_Error_Logger')) {
	457	+ try {
	458	+ Metasync_Error_Logger::log(
	459	+ Metasync_Error_Logger::CATEGORY_OTTO_RENDER,
	460	+ Metasync_Error_Logger::SEVERITY_ERROR,
	461	+ 'OTTO crawl-url job threw; retry policy applied',
	462	+ [
	463	+ 'operation' => 'crawl_url_job',
	464	+ 'outcome' => 'throwable',
	465	+ 'retry_count' => (int) $retry_count,
	466	+ 'exception' => get_class($e),
	467	+ ]
	468	+ );
	469	+ } catch (\Throwable $logger_error) {
	470	+ # Logging must never kill the handler after the retry is booked.
	471	+ error_log('MetaSync OTTO: crawl-url job error log failed: ' . $logger_error->getMessage()); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
	472	+ }
234	473	}
235	474	}
236	475	}
237	476
238	477	/**
239		- * Batch handler for cache warming and edge CDN purge.
	478	+ * Batch handler for the edge CDN purge.
240	479	*
241		- * Runs after all per-URL jobs have had time to complete. Batching these
242		- * operations avoids N separate API calls to edge providers (Cloudflare
243		- * supports up to 30 tags per request) and lets warm_urls() issue concurrent
244		- * fire-and-forget requests.
	480	+ * Runs after all per-URL jobs have had time to complete. Per-URL host caches
	481	+ * are cleared and re-warmed by the individual jobs; this batched event only
	482	+ * handles the edge CDN purge, which benefits from multi-URL API calls
	483	+ * (Cloudflare, Fastly, Akamai, Sucuri, Sevella, etc.).
245	484	*
246		- * @param array $routes List of fully-qualified URLs to warm and purge.
	485	+ * URLs still retrying are stragglers: the batch waits (up to
	486	+ * METASYNC_OTTO_BATCH_MAX_WAITS reschedules 60s apart) so their freshly
	487	+ * synced content — not stale content — reaches the edge. URLs that complete
	488	+ * after the wait budget purge the edge themselves from the retry path.
	489	+ *
	490	+ * @param array $routes List of fully-qualified URLs in this batch.
	491	+ * @param int $wait_count How many times this batch has already waited.
247	492	*/
248		-function metasync_handle_otto_batch_cache_job($routes) {
249		- try {
250		- # Re-populate the cache so OTTO-modified output is what gets stored.
251		- # Per-URL host cache was already cleared by individual jobs.
252		- # warm_urls() hits each URL with a non-blocking request so our code —
253		- # with fresh transients and post meta — is first to populate the host cache.
254		- Metasync_Cache_Purge::warm_urls($routes);
	493	+function metasync_handle_otto_batch_cache_job($routes = array(), $wait_count = 0) {
	494	+ // WP-Cron events persist independently of plugin code. A stale or malformed
	495	+ // event may therefore invoke this callback without its required route list.
	496	+ // Docblock type isn't enforced at runtime; a stale cron record can pass a non-array.
	497	+ // @phpstan-ignore-next-line function.alreadyNarrowedType
	498	+ if (!is_array($routes) || empty($routes)) {
	499	+ error_log('MetaSync OTTO: skipping batch cache job without valid routes.'); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
	500	+ return;
	501	+ }
255	502
256		- # Purge edge CDN caches (Cloudflare, Fastly, Akamai, Sucuri, Sevalla, etc.)
257		- # Tag-based providers purge only the affected posts; full-flush providers fire once per batch.
258		- Metasync_Edge_Cache_Purge::purge($routes);
259		- } catch (Exception $e) {
260		- metasync_record_failed_action('metasync_process_otto_batch_cache_job');
261		- error_log('MetaSync OTTO: batch cache job failed for ' . count($routes) . ' URLs: ' . $e->getMessage() . ' in ' . $e->getFile() . ':' . $e->getLine());
262		- }
263		-}
	503	+ $max_waits = defined('METASYNC_OTTO_BATCH_MAX_WAITS') ? METASYNC_OTTO_BATCH_MAX_WAITS : 3;
	504	+ // Docblock type isn't enforced at runtime; a stale cron record can pass a non-numeric.
	505	+ // @phpstan-ignore-next-line function.alreadyNarrowedType
	506	+ $wait_count = is_numeric($wait_count) ? (int) $wait_count : 0;
264	507
265		-# NOTE: Cache system removed - these functions are no longer needed
266		-# Kept for backward compatibility in case old cache directories need cleanup
267		-function metasync_deleteDir($dir) {
268		- if (!is_dir($dir)) {
269		- return false;
270		- }
271		- $files = array_diff(scandir($dir), array('.', '..'));
272		- foreach ($files as $file) {
273		- $filePath = $dir . DIRECTORY_SEPARATOR . $file;
274		- if (is_dir($filePath)) {
275		- metasync_deleteDir($filePath);
	508	+ # Split the batch: terminally-done URLs vs. URLs with work still in
	509	+ # flight. Unknown-state URLs (legacy events scheduled before the status
	510	+ # store existed, or deduped duplicates whose entry expired) count as done
	511	+ # — their per-URL job has either finished or never will, and both mean
	512	+ # the edge purge should not stall behind them.
	513	+ $ready = array();
	514	+ $stragglers = 0;
	515	+ foreach ($routes as $route) {
	516	+ $entry = Metasync_Otto_Job_Status::get($route);
	517	+ $state = is_array($entry) && isset($entry['state']) ? $entry['state'] : null;
	518	+ if ($state === Metasync_Otto_Job_Status::STATE_RETRYING
	519	+ || $state === Metasync_Otto_Job_Status::STATE_ACCEPTED) {
	520	+ $stragglers++;
276	521	} else {
277		- unlink($filePath);
	522	+ # completed / failed / unknown: nothing left to wait for.
	523	+ $ready[] = $route;
278	524	}
279	525	}
280		- return rmdir($dir);
281		-}
282	526
283		-# Cleanup function for removing old cache directories (if they exist)
284		-function metasync_invalidate_all_caches($folder = ''){
285		- # Cache system removed - this function only exists to clean up old cache directories
286		- if(!defined('WP_CONTENT_DIR')){
287		- return false;
	527	+ # Still live work in the batch and wait budget left: hold the edge purge.
	528	+ if ($stragglers > 0 && $wait_count < $max_waits) {
	529	+ metasync_otto_schedule_single_event(time() + 60, 'metasync_process_otto_batch_cache_job', array($routes, $wait_count + 1));
	530	+ return;
288	531	}
289		- $wp_content_dir = WP_CONTENT_DIR;
290		- $cache_dir = $wp_content_dir . '/metasync_caches';
291		- if(in_array($folder, ['posts', 'pages'])){
292		- $cache_dir = $cache_dir . '/' . $folder;
	532	+
	533	+ if (empty($ready)) {
	534	+ return;
293	535	}
294		- if(is_dir($cache_dir)){
295		- metasync_deleteDir($cache_dir);
	536	+
	537	+ try {
	538	+ # Purge edge CDN caches (Cloudflare, Fastly, Akamai, Sucuri, Sevalla, etc.)
	539	+ # Tag-based providers purge only the affected posts; full-flush providers fire once per batch.
	540	+ Metasync_Edge_Cache_Purge::purge($ready);
	541	+ } catch (Exception $e) {
	542	+ metasync_record_failed_action('metasync_process_otto_batch_cache_job');
	543	+ error_log('MetaSync OTTO: batch cache job failed for ' . count($ready) . ' URLs: ' . $e->getMessage() . ' in ' . $e->getFile() . ':' . $e->getLine()); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
296	544	}
297	545	}
298	546
299	547	/**
300		- * Determine whether a post is a MetaSync-built custom page that ships its own
301		- * complete, self-contained SEO (Custom HTML pages and LPS-imported pages).
	548	+ * Drainer: work off URLs that could not be scheduled inline (batch overflow
	549	+ * or a cron that refused the event) from the durable pending queue.
302	550	*
303		- * These pages already include a full SEO head (title, meta description, OG,
304		- * Twitter, JSON-LD schema) generated by LPS, so OTTO must not run on them —
305		- * otherwise it injects/overwrites SEO from a different/older OTTO project and
306		- * produces a page with two contradictory SEO identities (WP-440).
307		- *
308		- * @param int $post_id Post ID to inspect.
309		- * @return bool True when the post carries any custom-page / LPS marker.
	551	+ * Processes at most 25 URLs per run to stay friendly to WP-Cron, schedules
	552	+ * the same batched edge purge as the webhook path for the chunk it handled,
	553	+ * and reschedules itself while the queue is non-empty. URLs it takes off
	554	+ * the queue are scheduled unconditionally unless a real cron event already
	555	+ * exists for them — the webhook's overflow bookkeeping must never veto,
	556	+ * because by then the URL has already left the durable queue.
310	557	*/
311		-function metasync_is_custom_or_lps_page($post_id){
312		- $post_id = (int) $post_id;
313		- if ($post_id <= 0 || !class_exists('Metasync_Custom_Pages') || !defined('Metasync_Custom_Pages::META_LPS_IMPORT') || !defined('Metasync_Custom_Pages::META_IS_CUSTOM_HTML_PAGE') || !defined('Metasync_Custom_Pages::META_CREATED_VIA_API')) {
314		- return false;
	558	+function metasync_handle_otto_pending_drainer() {
	559	+ $chunk = Metasync_Otto_Job_Status::take_pending(25);
	560	+ if (empty($chunk)) {
	561	+ return;
315	562	}
316	563
317		- if (get_post_meta($post_id, Metasync_Custom_Pages::META_IS_CUSTOM_HTML_PAGE, true) === '1') {
318		- return true;
	564	+ $now = time();
	565	+ $scheduled = 0;
	566	+ foreach ($chunk as $key => $route) {
	567	+ # This run owns the chunk: take_pending() already removed these URLs
	568	+ # from the durable queue, so skipping here would drop them. Only a
	569	+ # real cron event vetoes — the webhook's overflow ledger row must
	570	+ # not (see metasync_otto_job_event_pending).
	571	+ if (metasync_otto_job_event_pending($route)) {
	572	+ continue;
	573	+ }
	574	+ if (metasync_otto_schedule_single_event($now + $key, 'metasync_process_otto_crawl_url_job', array($route))) {
	575	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'drain', 0, '');
	576	+ $scheduled++;
	577	+ } else {
	578	+ # Cron still refusing: put it back at the end of the queue.
	579	+ Metasync_Otto_Job_Status::enqueue_pending($route);
	580	+ }
319	581	}
320		- if (!empty(get_post_meta($post_id, Metasync_Custom_Pages::META_LPS_IMPORT, true))) {
321		- return true;
	582	+
	583	+ if ($scheduled > 0) {
	584	+ metasync_otto_schedule_single_event($now + count($chunk) + 5, 'metasync_process_otto_batch_cache_job', array($chunk));
322	585	}
323		- if (!empty(get_post_meta($post_id, Metasync_Custom_Pages::META_CREATED_VIA_API, true))) {
324		- return true;
	586	+
	587	+ if (Metasync_Otto_Job_Status::pending_count() > 0) {
	588	+ metasync_otto_schedule_single_event($now + 60, 'metasync_otto_pending_drainer', array());
325	589	}
	590	+}
326	591
327		- return false;
328		-}
	592	+// metasync_is_custom_or_lps_page() now lives in includes/metasync-helpers.php
	593	+// (loaded unconditionally before this file) so all SEO surfaces share one rule.
329	594
330	595	/**
331	596	* Detect an Elementor editor / preview request that OTTO must never process.
332	597	*
		@@ -347,15 +612,15 @@
347	612	*/
348	613	function metasync_is_elementor_editor_request() {
349	614	# The preview iframe loads the front end with ?elementor-preview=POST_ID.
350	615	# This is the request OTTO would otherwise buffer and corrupt.
351		- if (isset($_GET['elementor-preview'])) {
	616	+ if (isset($_GET['elementor-preview'])) { // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only editor request detection
352	617	return true;
353	618	}
354	619
355	620	# Elementor editor / app entry points carried as an action on the front end.
356		- if (isset($_REQUEST['action'])
357		- && in_array($_REQUEST['action'], array('elementor', 'elementor_ajax'), true)
	621	+ if (isset($_REQUEST['action']) // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only editor request detection
	622	+ && in_array($_REQUEST['action'], array('elementor', 'elementor_ajax'), true) // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only editor request detection
358	623	) {
359	624	return true;
360	625	}
361	626
		@@ -384,9 +649,9 @@
384	649	# SSR is SKIPPED for: cart, checkout
385	650	# Note: title/description filters (pre_get_document_title, wp_head meta desc)
386	651	# run on ALL pages regardless — they are hooked unconditionally in seo-functions.php
387	652
388		- # ── WooCommerce-independent cart/checkout protection (WP-374) ──────────
	653	+ # ── WooCommerce-independent cart/checkout protection ──────────
389	654	# The is_cart()/is_checkout() guards below only recognize WooCommerce. Carts
390	655	# from other systems — e.g. the Point of Rental "Catalog" plugin on
391	656	# venturarental.com — are invisible to them, so OTTO would process those
392	657	# pages and let caching layers store them. A cached cart page corrupts the
		@@ -399,8 +664,24 @@
399	664	if (($_SERVER['REQUEST_METHOD'] ?? 'GET') !== 'GET') {
400	665	return;
401	666	}
402	667
	668	+ # Never run OTTO on WordPress core draft/pending previews
	669	+ # (?p=<ID>&preview=true, with preview_id/preview_nonce on the full link).
	670	+ # Previews are per-user, non-cacheable, and routed purely by query string —
	671	+ # and OTTO's route identity ignores the query string, so a processed preview
	672	+ # is answered with the route's public page (the homepage for /?p=…) instead
	673	+ # of the draft the editor asked for. Skip early and let WP render it.
	674	+ if (
	675	+ (isset($_GET['preview']) && $_GET['preview'] === 'true') || // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
	676	+ !empty($_GET['preview_id']) // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
	677	+ ) {
	678	+ if (!defined('DONOTCACHEPAGE')) {
	679	+ define('DONOTCACHEPAGE', true);
	680	+ }
	681	+ return;
	682	+ }
	683	+
403	684	# Never run OTTO on the Elementor editor/preview iframe. It is a logged-in
404	685	# front-end GET (not is_admin()), so it slips past the admin guard below;
405	686	# buffering and rewriting it breaks the editor canvas ("fails to load").
406	687	if (metasync_is_elementor_editor_request()) {
		@@ -406,8 +687,24 @@
406	687	if (metasync_is_elementor_editor_request()) {
407	688	return;
408	689	}
409	690
	691	+ # WP core file-editor self-check. When an admin saves a THEME file
	692	+ # via the Plugin/Theme Editor, wp_edit_theme_plugin_file() fires an internal
	693	+ # loopback GET to home_url('/') carrying wp_scrape_key/wp_scrape_nonce to
	694	+ # detect a white-screen. That request is NOT is_admin(), so — exactly like
	695	+ # the Elementor case above — it slips past the admin/AJAX/REST guards below.
	696	+ # OTTO must not output-buffer or rewrite it: the SimpleHtmlDom pass (and any
	697	+ # fatal thrown inside the ob_start() callback, e.g. under memory pressure)
	698	+ # corrupts the scrape and surfaces as the misleading
	699	+ # "preg_match(): Cannot use output buffering in output buffering display
	700	+ # handlers" fatal. Skipping OTTO lets WP render the page normally so the
	701	+ # scrape works. Returning here also avoids the HTTP-fallback path firing a
	702	+ # second loopback for an already-internal self-request.
	703	+ if (metasync_is_scrape_request()) {
	704	+ return;
	705	+ }
	706	+
410	707	# Skip OTTO on cart/checkout-type paths regardless of the cart plugin, and
411	708	# signal caching plugins (WP Rocket/Kinsta/etc.) to never cache them — a
412	709	# cached cart/checkout page is precisely what corrupts cart state.
413	710	# Matched against the FIRST path segment (relative to the WP home path) so
		@@ -417,9 +714,9 @@
417	714	# then lower-case and strip surrounding slashes.
418	715	$otto_req_path = explode('?', (string) wp_unslash($_SERVER['REQUEST_URI'] ?? ''), 2)[0];
419	716	$otto_req_path = strtolower(trim($otto_req_path, '/'));
420	717	# Strip the site's base path so first-segment matching works on subdir installs.
421		- $otto_home_path = trim((string) parse_url(home_url(), PHP_URL_PATH), '/');
	718	+ $otto_home_path = trim((string) wp_parse_url(home_url(), PHP_URL_PATH), '/');
422	719	if ($otto_home_path !== '' && strpos($otto_req_path, $otto_home_path . '/') === 0) {
423	720	$otto_req_path = substr($otto_req_path, strlen($otto_home_path) + 1);
424	721	}
425	722	$otto_first_segment = explode('/', $otto_req_path, 2)[0];
		@@ -437,11 +734,11 @@
437	734	# ──────────────────────────────────────────────────────────────────────
438	735
439	736	if (
440	737	# disable ajax calls
441		- isset($_GET['ucfrontajaxaction']) ||
	738	+ isset($_GET['ucfrontajaxaction']) || // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
442	739	# OTTO Preview mode - skip OTTO when previewing original content
443		- (isset($_GET['otto_preview']) && $_GET['otto_preview'] === '1') ||
	740	+ (isset($_GET['otto_preview']) && $_GET['otto_preview'] === '1') || // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
444	741	# WooCommerce shop archive page only (products and categories now use SSR)
445	742	//(function_exists('is_shop') && is_shop()) ||
446	743	# Cart page
447	744	(function_exists('is_cart') && is_cart()) ||
		@@ -453,32 +750,32 @@
453	750	(function_exists('wp_doing_ajax') && wp_doing_ajax()) ||
454	751	# check by constant
455	752	(defined('DOING_AJAX') && DOING_AJAX) ||
456	753	# WooCommerce AJAX endpoint (e.g., ?wc-ajax=update_cart)
457		- (isset($_REQUEST['wc-ajax']) && !empty($_REQUEST['wc-ajax'])) ||
	754	+ (isset($_REQUEST['wc-ajax']) && !empty($_REQUEST['wc-ajax'])) || // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
458	755	# AJAX requests via X-Requested-With header
459	756	(isset($_SERVER['HTTP_X_REQUESTED_WITH']) && strtolower($_SERVER['HTTP_X_REQUESTED_WITH']) === 'xmlhttprequest') ||
460	757	# Gravity Forms submission detection - skip OTTO to allow form processing
461		- (isset($_POST['gform_submit']) && (
462		- is_array($_POST['gform_submit']) ||
463		- (is_string($_POST['gform_submit']) && isset($_POST['is_submit_' . $_POST['gform_submit']]) && !empty($_POST['gform_submit']))
	758	+ (isset($_POST['gform_submit']) && ( // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
	759	+ is_array($_POST['gform_submit']) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
	760	+ (is_string($_POST['gform_submit']) && isset($_POST['is_submit_' . $_POST['gform_submit']]) && !empty($_POST['gform_submit'])) // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
464	761	)) ||
465	762	# Gravity Forms AJAX submission
466		- (isset($_POST['gform_ajax']) && isset($_POST['gform_submit']) && (
467		- is_array($_POST['gform_submit']) || !empty($_POST['gform_submit'])
	763	+ (isset($_POST['gform_ajax']) && isset($_POST['gform_submit']) && ( // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
	764	+ is_array($_POST['gform_submit']) || !empty($_POST['gform_submit']) // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
468	765	)) ||
469	766	# Gravity Forms file upload
470		- (isset($_POST['gform_uploaded_files'])) ||
	767	+ (isset($_POST['gform_uploaded_files'])) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
471	768	# Any Gravity Forms POST parameter
472		- (isset($_POST['gform_submit']) || isset($_POST['gform_unique_id']) || isset($_POST['gform_field_values'])) ||
	769	+ (isset($_POST['gform_submit']) || isset($_POST['gform_unique_id']) || isset($_POST['gform_field_values'])) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
473	770	# Formidable Forms AJAX submission detection - skip OTTO to allow form processing
474		- (isset($_POST['action']) && $_POST['action'] === 'frm_entries_create') ||
	771	+ (isset($_POST['action']) && $_POST['action'] === 'frm_entries_create') || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
475	772	# Formidable Forms POST parameters
476		- (isset($_POST['form_id']) && !empty($_POST['form_id'])) ||
	773	+ (isset($_POST['form_id']) && !empty($_POST['form_id'])) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
477	774	# Formidable Forms action parameter
478		- (isset($_POST['frm_action']) && !empty($_POST['frm_action'])) ||
	775	+ (isset($_POST['frm_action']) && !empty($_POST['frm_action'])) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
479	776	# Formidable Forms item_key (used in form submissions)
480		- (isset($_POST['item_key']) && !empty($_POST['item_key']))
	777	+ (isset($_POST['item_key']) && !empty($_POST['item_key'])) // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
481	778	) {
482	779	return;
483	780	}
484	781
		@@ -512,14 +809,34 @@
512	809	# suggestions for machine-readable XML, and the upstream API returns
513	810	# API_ERROR for these URLs — which then stamps misleading
514	811	# X-MetaSync-OTTO-Cache: API_ERROR / X-MetaSync-OTTO-Method: NONE headers
515	812	# onto otherwise-healthy sitemap responses and causes false-alarm bug
516		- # reports from customers. See WP-353.
	813	+ # reports from customers.
517	814	$request_path = strtok($request_uri, '?');
518	815	if ($request_path && preg_match('#\.xml$#i', $request_path)) {
519	816	return;
520	817	}
521	818
	819	+ # Machine-readable core endpoints. Using WordPress's own conditionals rather
	820	+ # than matching the path: they are exact, and a path pattern here would also
	821	+ # match ordinary content whose slug merely starts with the same characters
	822	+ # (e.g. /robots.txt-explained/, /favicon.ico-vs-svg/).
	823	+ # is_favicon() is WP 5.4+; this plugin supports 5.2.
	824	+ if (is_robots() || (function_exists('is_favicon') && is_favicon())) {
	825	+ return;
	826	+ }
	827	+
	828	+ # Protocol paths and static assets that are never OTTO-able content.
	829	+ # The guard fails open — if the strategy is unavailable we let the request
	830	+ # proceed rather than fatal. metasync_otto_class_provides() rather than a
	831	+ # bare class_exists(): a partially updated install can leave an older copy of
	832	+ # the strategy class loaded that predates is_non_content_path(), which
	833	+ # satisfies class_exists() and then fatals on every front-end request.
	834	+ if (metasync_otto_class_provides('Metasync_Otto_Render_Strategy', 'is_non_content_path')
	835	+ && Metasync_Otto_Render_Strategy::is_non_content_path($request_path)) {
	836	+ return;
	837	+ }
	838	+
522	839	# BOT DETECTION: Always detect bots so crawl data reaches the SA backend.
523	840	# Real bots don't execute JS, so otto-tracker.js never fires for them.
524	841	# push_crawl_log_to_sa() fires a non-blocking wp_remote_post here instead.
525	842	$bot_detector = Metasync_Otto_Bot_Detector::get_instance();
		@@ -527,15 +844,46 @@
527	844	if ( $detection['is_bot'] ) {
528	845	$bot_detector->push_crawl_log_to_sa( $detection, $current_url );
529	846	}
530	847
531		- # Throttle OTTO rendering for non-search-engine bots: at most one render
532		- # per URL+bot every 5 minutes. Humans and verified search engines
533		- # (Googlebot, Bingbot, etc.) are never throttled — their hits always get
534		- # full OTTO content so indexing remains current. SEO tools, AI scrapers,
535		- # uptime monitors, and unverified crawlers are de-duplicated so repeat
536		- # hits don't re-trigger the expensive DOM-rewriting path.
537		- if ( $detection['is_bot'] && $detection['bot_type'] !== 'search_engine' ) {
	848	+ # A headless frontend renders the public HTML, so stop after the cheap cache
	849	+ # protections and crawl telemetry, before any OTTO rendering work begins.
	850	+ #
	851	+ # Routed through metasync_headless_is_active() rather than calling the class
	852	+ # directly: Metasync_Headless_Config resolves only through the committed
	853	+ # Composer classmap, so a partially updated install can leave this newer
	854	+ # caller beside a missing or older copy with no fallback lookup path. This
	855	+ # function is hooked to `wp`, so an unguarded static call is a white screen
	856	+ # on every front-end page view. The helper treats an unresolvable class as
	857	+ # "not headless", which keeps OTTO behaving as it did before headless mode.
	858	+ # @phpstan-ignore-next-line function.alreadyNarrowedType
	859	+ $metasync_is_headless = function_exists('metasync_headless_is_active')
	860	+ ? metasync_headless_is_active()
	861	+ : (class_exists('Metasync_Headless_Config') && Metasync_Headless_Config::is_active());
	862	+
	863	+ if ($metasync_is_headless) {
	864	+ return;
	865	+ }
	866	+
	867	+ # Throttle OTTO rendering for bots that are not exempt: at most one render
	868	+ # per URL+bot every 5 minutes. Humans, search engines (Googlebot, Bingbot,
	869	+ # Yahoo's Slurp, Sogou, Exabot, Applebot, ...), host cache warmers and
	870	+ # synthetic performance auditors are never throttled — their hits always get
	871	+ # full OTTO content, so indexing stays current, the page cache can actually
	872	+ # be filled, and audit tools measure what a real visitor sees. SEO tools,
	873	+ # social crawlers, archivers, AI scrapers, uptime monitors and unrecognized
	874	+ # crawlers are de-duplicated so repeat hits don't re-trigger the expensive
	875	+ # DOM-rewriting path.
	876	+ #
	877	+ # Two complementary gates, deliberately kept separate:
	878	+ # - is_unthrottled_category() consults the classifier's explicit per-bot
	879	+ # category map, so the exempt set is defined in exactly one place.
	880	+ # - metasync_otto_is_unthrottled_infrastructure_agent() matches the raw
	881	+ # user-agent (plus a query marker and a site filter), which catches
	882	+ # preload agents that have no entry in the map at all.
	883	+ if ( $detection['is_bot']
	884	+ && ! Metasync_Otto_Bot_Detector::is_unthrottled_category( $detection['bot_type'] )
	885	+ && ! metasync_otto_is_unthrottled_infrastructure_agent( $detection ) ) {
538	886	$normalized_url = strtok( $current_url, '?' );
539	887	if ( $normalized_url === false ) {
540	888	$normalized_url = $current_url;
541	889	}
		@@ -554,8 +902,27 @@
554	902	}
555	903	set_transient( $render_throttle_key, 1, 5 * MINUTE_IN_SECONDS );
556	904	}
557	905
	906	+ # Skip OTTO on 404s.
	907	+ #
	908	+ # OTTO suggestions are keyed to real, crawled pages. Asking the API about a URL
	909	+ # that returns 404 spends a full API_TIMEOUT of PHP-FPM worker time on a
	910	+ # guaranteed-empty answer. Not theoretical: in one production sample, 21 of 54
	911	+ # crawler requests were 404s on URLs that never existed on the site, each
	912	+ # costing ~2s of a worker.
	913	+ #
	914	+ # Deliberately placed AFTER bot detection so push_crawl_log_to_sa() still
	915	+ # reports the crawl to SearchAtlas — that telemetry is how the crawler side
	916	+ # can learn to stop requesting these URLs. Only the expensive suggestions
	917	+ # lookup is skipped.
	918	+ #
	919	+ # Filterable because "should a 404 page get OTTO treatment" is arguably a site
	920	+ # decision, even though there is no known case where it should.
	921	+ if (is_404() && apply_filters('metasync_otto_skip_on_404', true)) {
	922	+ return;
	923	+ }
	924	+
558	925	# Optionally skip OTTO processing for bot traffic (when the setting is enabled)
559	926	if ($bot_detector->should_skip_otto()) {
560	927	// Log the bot locally and count the saved API call
561	928	$bot_detector->log_detection($detection);
		@@ -566,9 +933,9 @@
566	933	return;
567	934	}
568	935
569	936	# OPTIMIZED: Check if Otto should be disabled for WP Rocket compatibility
570		- if (class_exists('WP_Rocket')) {
	937	+ if (defined('WP_ROCKET_VERSION') || class_exists('WP_Rocket')) {
571	938	$wp_rocket_compat_mode = Metasync_Otto_Config::get_wp_rocket_compat_mode();
572	939
573	940	if ($wp_rocket_compat_mode === 'disable_otto') {
574	941	return; # Exit early, Otto is disabled when WP Rocket is active
		@@ -574,22 +941,55 @@
574	941	return; # Exit early, Otto is disabled when WP Rocket is active
575	942	}
576	943	}
577	944
578		- # Skip OTTO for Divi AJAX pagination and paginated archive requests (WP-315).
579		- # ?et_blog = Divi AJAX pagination callback
	945	+ # Skip OTTO for Divi AJAX pagination, and for paginated archive requests
	946	+ # on Divi-built sites.
	947	+ # ?et_blog = Divi AJAX pagination callback — always skipped, on every
	948	+ # theme, because Divi fires it regardless of the site layout.
580	949	# /page/N/ = paginated blog/archive pages — OTTO's buffer/HTTP render causes
581	950	# module numbering mismatch between page 1 (with TB template) and page N
582	951	# (without TB template), breaking Divi's JS pagination selector matching.
583		- if (isset($_GET['et_blog']) || (is_paged() && !is_singular())) {
	952	+ # That mismatch is a Divi rendering bug, so the /page/N/ skip is gated on
	953	+ # Divi detection: non-Divi sites (Elementor + the7, ...) keep OTTO on real
	954	+ # paginated archives. Detection covers the Divi theme and its child themes
	955	+ # (get_template() returns the PARENT template dir, "Divi" either way), the
	956	+ # Extra theme and the Divi Builder plugin (both define ET_BUILDER_VERSION
	957	+ # / et_setup_theme — same TB machinery, same bug).
	958	+ $is_divi = defined('ET_BUILDER_VERSION')
	959	+ || function_exists('et_setup_theme')
	960	+ || wp_get_theme()->get_template() === 'Divi';
	961	+ if (isset($_GET['et_blog']) || ($is_divi && is_paged() && !is_singular())) { // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
584	962	return;
585	963	}
586	964
	965	+ # Skip OTTO on search-results pages. A search request carries its
	966	+ # meaning entirely in the query string (e.g. /?s=term, or FiboSearch's
	967	+ # /?s=term&post_type=product&dgwt_wcas=1). get_route() intentionally strips
	968	+ # the query string to build a canonical, cache-key-stable route — which
	969	+ # collapses a query-only search URL to the site root ('/'). OTTO then finds
	970	+ # the HOME page's suggestions for that route and, on the HTTP render path,
	971	+ # serves the home page in place of the search results (the address bar keeps
	972	+ # the search URL — it is a silent content swap, not a redirect). This only
	973	+ # surfaces for logged-in / uncached visitors, since page caches serve the
	974	+ # correct pre-rendered search page to everyone else. Search-results pages
	975	+ # have no per-canonical-URL OTTO suggestions of their own, so OTTO must never
	976	+ # run here. The raw $_GET['s'] check is a defensive fallback for setups where
	977	+ # the main query is altered before this point.
	978	+ if (is_search() || !empty($_GET['s'])) { // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
	979	+ return;
	980	+ }
	981	+
587	982	# Handle cache plugin compatibility early - before any caching happens
588	983	metasync_otto_handle_cache_compatibility();
589	984
590	985	# check if OTTO is disabled for this specific page/post
591		- $post_id = get_the_ID();
	986	+ # Gate on is_singular() for the same reason the custom/LPS check below does:
	987	+ # on an archive, search or term query get_the_ID() returns the first post of
	988	+ # the loop — WP::register_globals() seeds $GLOBALS['post'] with it — not the
	989	+ # page being rendered. Without the gate, one excluded or per-page-disabled
	990	+ # post silently takes OTTO off every archive that happens to list it first.
	991	+ $post_id = is_singular() ? ( get_queried_object_id() ?: get_the_ID() ) : 0;
592	992	if ($post_id && class_exists('Metasync_Otto_Frontend_Toolbar')) {
593	993	if (Metasync_Otto_Frontend_Toolbar::is_otto_disabled($post_id)) {
594	994	return;
595	995	}
		@@ -596,9 +996,9 @@
596	996	}
597	997
598	998	# Skip OTTO on MetaSync custom HTML / LPS-imported pages — they ship their
599	999	# own complete, self-contained SEO and OTTO must not inject or overwrite it
600		- # with a different/older project's SEO (WP-440). Resolve the queried object
	1000	+ # with a different/older project's SEO. Resolve the queried object
601	1001	# id (with get_the_ID() fallback) so the static-front-page case — an LPS home
602	1002	# set as the WP front page, where is_page() is false — is still detected.
603	1003	# Applying the skip here, upstream of the single render_route_html() entry,
604	1004	# covers all three render paths (Rocket buffer, output buffer, HTTP fallback).
		@@ -603,12 +1003,12 @@
603	1003	# Applying the skip here, upstream of the single render_route_html() entry,
604	1004	# covers all three render paths (Rocket buffer, output buffer, HTTP fallback).
605	1005	# Only singular views (posts/pages, incl. a static front page) can be a
606	1006	# custom/LPS page; gate on is_singular() so an archive/search/term query
607		- # can never have its object id mistaken for a custom page's post id (WP-440).
	1007	+ # can never have its object id mistaken for a custom page's post id.
608	1008	$custom_page_id = is_singular() ? ( get_queried_object_id() ?: get_the_ID() ) : 0;
609	1009	if (metasync_is_custom_or_lps_page($custom_page_id)) {
610		- if (!headers_sent()) {
	1010	+ if (!headers_sent() && Metasync_Otto_Render_Strategy::diagnostics_enabled()) {
611	1011	header('X-MetaSync-OTTO-Method: EXCLUDED');
612	1012	}
613	1013	return;
614	1014	}
		@@ -613,15 +1013,32 @@
613	1013	return;
614	1014	}
615	1015
616	1016	# check if we are having an otto request
617		- if(!empty($_GET['is_otto_page_fetch'])){
	1017	+ #
	1018	+ # Recognise our own internal fetch by header/User-Agent as well as by the
	1019	+ # query parameter. A redirect that rebuilds the URL strips the parameter,
	1020	+ # and without this the redirected request re-enters OTTO and fires another
	1021	+ # internal fetch - an unbounded self-request loop.
	1022	+ # method_exists() is redundant to static analysis - this MR adds the method,
	1023	+ # so PHPStan proves the call always true. It is kept for the upgrade window:
	1024	+ # during a plugin update an opcache can still hold the previous
	1025	+ # Metasync_Otto_Render_Strategy, where class_exists() passes but the method
	1026	+ # is absent, and an unguarded static call would fatal the page.
	1027	+ # @phpstan-ignore-next-line function.alreadyNarrowedType
	1028	+ $otto_has_fetch_detector = class_exists('Metasync_Otto_Render_Strategy') && method_exists('Metasync_Otto_Render_Strategy', 'is_internal_fetch');
618	1029
	1030	+ $otto_is_internal_fetch = $otto_has_fetch_detector
	1031	+ ? Metasync_Otto_Render_Strategy::is_internal_fetch()
	1032	+ : !empty($_GET['is_otto_page_fetch']); // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only loopback marker, request-scoped filters only
	1033	+
	1034	+ if($otto_is_internal_fetch){
	1035	+
619	1036	# Block SEO plugins NOW for this internal fetch request
620	1037	# metasync_otto_block_seo_plugins();
621	1038	# $_SERVER['REQUEST_URI'] = remove_query_arg('is_otto_page_fetch', $_SERVER['REQUEST_URI']);
622		- $block_title = !empty($_GET['otto_block_title']) && $_GET['otto_block_title'] === '1';
623		- $block_description = !empty($_GET['otto_block_desc']) && $_GET['otto_block_desc'] === '1';
	1039	+ $block_title = !empty($_GET['otto_block_title']) && $_GET['otto_block_title'] === '1'; // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only loopback marker, request-scoped filters only
	1040	+ $block_description = !empty($_GET['otto_block_desc']) && $_GET['otto_block_desc'] === '1'; // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only loopback marker, request-scoped filters only
624	1041
625	1042	# Block SEO plugins conditionally based on what Otto has
626	1043	if ($block_title || $block_description) {
627	1044	metasync_otto_block_seo_plugins($block_title, $block_description);
		@@ -657,8 +1074,19 @@
657	1074
658	1075	# get the otto uuid
659	1076	$otto_uuid = Metasync_Otto_Config::get_otto_uuid();
660	1077
	1078	+ # Upgrade-window guard: the bundled DOM parser resolves only through the
	1079	+ # otto/vendor autoloader required at the top of this file. While an update
	1080	+ # is mid-copy (or a stale opcache classmap is still resident), that
	1081	+ # dependency can briefly be undefined here, and constructing the render
	1082	+ # stack below would throw a class-not-found Error and take the whole page
	1083	+ # down with it. Degrade to an OTTO-less page for this request instead —
	1084	+ # the next request, on the settled tree, renders normally.
	1085	+ if ( ! class_exists( 'simplehtmldom\HtmlDocument' ) ) {
	1086	+ return;
	1087	+ }
	1088	+
661	1089	# start the class
662	1090	$otto = new Metasync_otto_pixel($otto_uuid);
663	1091
664	1092	# call render
		@@ -671,12 +1099,12 @@
671	1099	* This function is called early in the WordPress lifecycle
672	1100	*/
673	1101	function metasync_otto_handle_cache_compatibility() {
674	1102	# Detect active plugins. SG Optimizer detection now lives in
675		- # metasync_otto_disable_sg_page_cache() (WP-498), which is invoked later
	1103	+ # metasync_otto_disable_sg_page_cache(), which is invoked later
676	1104	# once OTTO confirms suggestions for the URL.
677	1105	$brizy_active = class_exists('Brizy_Editor') || defined('BRIZY_VERSION');
678		- $wp_rocket_active = class_exists('WP_Rocket');
	1106	+ $wp_rocket_active = defined('WP_ROCKET_VERSION') || class_exists('WP_Rocket');
679	1107
680	1108	# OPTIMIZED: Get configuration option using cached config
681	1109	$wp_rocket_compat_mode = Metasync_Otto_Config::get_wp_rocket_compat_mode();
682	1110
		@@ -690,9 +1118,9 @@
690	1118	if ($cached !== false) {
691	1119	$has_brizy_posts = ($cached === 'yes');
692	1120	} else {
693	1121	# Query database only if cache missed
694		- $has_brizy_posts = $wpdb->get_var(
	1122	+ $has_brizy_posts = $wpdb->get_var( // phpcs:ignore WordPress.DB.DirectDatabaseQuery -- feature-detection probe for Brizy posts — postmeta flag scan has no WordPress API
695	1123	"SELECT COUNT(*) FROM {$wpdb->postmeta}
696	1124	WHERE meta_key = 'brizy_post_uid'
697	1125	AND meta_value != ''
698	1126	LIMIT 1"
		@@ -722,9 +1150,9 @@
722	1150
723	1151	# Case 3: WP Rocket active with auto/buffer mode - DON'T set DONOTCACHEPAGE
724	1152	# This allows WP Rocket optimizations to continue working
725	1153
726		- # NOTE (WP-498): SiteGround SG Optimizer cache bypass is intentionally NOT
	1154	+ # NOTE: SiteGround SG Optimizer cache bypass is intentionally NOT
727	1155	# handled here. Emitting no-cache headers / disabling SG cache on this hook
728	1156	# fired on EVERY front-end page — including pages OTTO never modifies —
729	1157	# forcing SG to bypass its page cache site-wide and spiking CPU. The SG
730	1158	# bypass is now deferred to metasync_otto_disable_sg_page_cache(), called
		@@ -743,15 +1171,15 @@
743	1171	* Called from Metasync_otto_pixel::render_route_html() once OTTO has confirmed
744	1172	* it has suggestions to apply to the requested URL. Scoping the no-cache
745	1173	* override to pages OTTO actually modifies — rather than emitting it on every
746	1174	* front-end page via the unconditional `wp` hook — keeps SG's page cache
747		- * working site-wide and avoids the CPU spike reported in WP-498.
	1175	+ * working site-wide and avoids the CPU spike reported in.
748	1176	*
749	1177	* Only meaningful on SiteGround sites without WP Rocket; a no-op otherwise.
750	1178	*/
751	1179	function metasync_otto_disable_sg_page_cache() {
752	1180	# Only relevant when SG Optimizer is active and WP Rocket is not present.
753		- if (class_exists('WP_Rocket')) {
	1181	+ if (defined('WP_ROCKET_VERSION') || class_exists('WP_Rocket')) {
754	1182	return;
755	1183	}
756	1184	if (!is_plugin_active('sg-cachepress/sg-cachepress.php')) {
757	1185	return;
		@@ -775,8 +1203,314 @@
775	1203	}
776	1204	}
777	1205
778	1206	/**
	1207	+ * How long a response that OTTO could not optimize may be cached, in seconds.
	1208	+ *
	1209	+ * Deliberately short rather than zero — see
	1210	+ * metasync_otto_limit_response_cache() for why.
	1211	+ */
	1212	+if (!defined('METASYNC_OTTO_FALLBACK_CACHE_TTL')) {
	1213	+ define('METASYNC_OTTO_FALLBACK_CACHE_TTL', 60);
	1214	+}
	1215	+
	1216	+/**
	1217	+ * Cap how long the current response may be cached, because OTTO could not apply
	1218	+ * its suggestions to it.
	1219	+ *
	1220	+ * OTTO fails open: when suggestions are unavailable, or when the HTML rewrite
	1221	+ * fails, the site's own un-optimized markup is served instead. That markup is
	1222	+ * valid — the harm is that a caching layer then stores it and serves it to
	1223	+ * everyone for the rest of its TTL (a host edge cache with s-maxage=86400 pins
	1224	+ * it for a day), which surfaces to customers as the OTTO title "reverting".
	1225	+ *
	1226	+ * A short TTL is used rather than `no-store`, which is the obvious choice but the
	1227	+ * wrong one. Some failures look transient and are not: a page that always breaks
	1228	+ * the rewriter, or a host that cannot fetch its own pages, fails identically on
	1229	+ * every request. `no-store` would make those permanently uncacheable, sending
	1230	+ * 100% of their traffic to the origin forever in exchange for markup that will
	1231	+ * never improve. A one-minute ceiling fixes the reported bug just as completely —
	1232	+ * a day-long pin becomes a minute — while capping the cost of being wrong about
	1233	+ * which failures recover.
	1234	+ *
	1235	+ * Different layers read different signals, so all of them are emitted:
	1236	+ *
	1237	+ * - Cache-Control CDNs, host edge caches, browsers. `max-age=0` keeps
	1238	+ * browsers revalidating while `s-maxage` lets shared
	1239	+ * caches absorb a burst for the ceiling's duration.
	1240	+ * - X-Accel-Expires Nginx's own directive. Evaluated with higher
	1241	+ * precedence than Cache-Control, and not covered by
	1242	+ * the `fastcgi_ignore_headers Cache-Control Expires`
	1243	+ * recipe common on managed hosts — so it can survive
	1244	+ * a config that discards Cache-Control. Nginx strips
	1245	+ * X-Accel-* before the response reaches the client,
	1246	+ * so there is no visitor-facing side effect.
	1247	+ * - Surrogate-Control Varnish and Fastly. Note that on stock Varnish the
	1248	+ * mere presence of this header suppresses its
	1249	+ * Cache-Control check, so it must carry a real
	1250	+ * lifetime rather than be left at some other value.
	1251	+ * - DONOTCACHEPAGE WordPress-level page caches (WP Rocket, W3TC,
	1252	+ * LiteSpeed, SG Optimizer). These run inside PHP and
	1253	+ * never see response headers. They have no notion of a
	1254	+ * short TTL — it is cache or do not — so they skip
	1255	+ * entirely. That is safe here because callers only
	1256	+ * invoke this for failures expected to resolve; a
	1257	+ * permanent one (an oversized page, a host that cannot
	1258	+ * reach itself) is left fully cacheable by the caller
	1259	+ * and never reaches this function.
	1260	+ *
	1261	+ * Cache-tagging headers (Surrogate-Key, Cache-Tag, Edge-Cache-Tag) are
	1262	+ * deliberately left in place. They are purge-targeting labels, not cacheability
	1263	+ * directives — no CDN stores a response *because* a tag is present. On a layer
	1264	+ * that stores this response anyway, those tags are the only handle a later purge
	1265	+ * has on it, so removing them would discard the recovery path exactly where it
	1266	+ * is needed most.
	1267	+ *
	1268	+ * Some layers may honour none of this; a managed host that overrides origin
	1269	+ * cache headers will cache the response regardless. This bounds the damage where
	1270	+ * it can and is inert where it cannot.
	1271	+ *
	1272	+ * @param string $reason Short machine-readable reason, surfaced as a response
	1273	+ * header for debugging (e.g. 'API_ERROR', 'RENDER_FAILED').
	1274	+ */
	1275	+/**
	1276	+ * Build the response headers that cap how long an un-optimized response may be
	1277	+ * cached.
	1278	+ *
	1279	+ * Thin wrapper over Metasync_Otto_Render_Strategy::fallback_cache_headers(), which
	1280	+ * owns the header set so it can be unit-tested (this file cannot be loaded in a
	1281	+ * test context). Returns an empty array if the strategy class is somehow absent,
	1282	+ * so a caller emits nothing rather than an untested divergent copy.
	1283	+ *
	1284	+ * @param int $ttl Cache ceiling in seconds.
	1285	+ * @param string $reason Optional machine-readable reason for the debug header.
	1286	+ * @return array Header name => value.
	1287	+ */
	1288	+function metasync_otto_fallback_cache_headers($ttl, $reason = '') {
	1289	+ if (!class_exists('Metasync_Otto_Render_Strategy')) {
	1290	+ return [];
	1291	+ }
	1292	+
	1293	+ return Metasync_Otto_Render_Strategy::fallback_cache_headers_for_response($ttl, $reason);
	1294	+}
	1295	+
	1296	+function metasync_otto_limit_response_cache($reason = '') {
	1297	+ # This function must only ever make the response LESS cacheable. header()
	1298	+ # replaces by default and the capped set carries X-Accel-Expires, which nginx
	1299	+ # evaluates ahead of Cache-Control — so emitting it unconditionally could
	1300	+ # override an upstream no-store and grant edge caching that nothing had
	1301	+ # granted. The strategy reconciles existing cache directives before emitting
	1302	+ # a capped set.
	1303	+
	1304	+ $strategy_available = class_exists('Metasync_Otto_Render_Strategy');
	1305	+
	1306	+ # 1. Logged-in users. OTTO renders for them unless the "disable on logged in"
	1307	+ # setting is switched on, which is off by default — so this is the common
	1308	+ # case, not an edge case. Their page may carry session-specific or
	1309	+ # membership-gated content, and capping at 60 seconds would still let a
	1310	+ # shared cache store and re-serve it to somebody else.
	1311	+ if (function_exists('is_user_logged_in') && is_user_logged_in()) {
	1312	+ if (!headers_sent() && $strategy_available) {
	1313	+ foreach (Metasync_Otto_Render_Strategy::no_store_headers($reason) as $name => $value) {
	1314	+ header($name . ': ' . $value);
	1315	+ }
	1316	+ }
	1317	+ if (!defined('DONOTCACHEPAGE')) {
	1318	+ define('DONOTCACHEPAGE', true);
	1319	+ }
	1320	+ return;
	1321	+ }
	1322	+
	1323	+ # 2. Something upstream already restricted this response. Leave its headers
	1324	+ # alone — a page deliberately marked no-store must not be handed a 60
	1325	+ # second shared-cache lifetime by us.
	1326	+ # The strategy turns an existing no-store into a consistent no-store set and
	1327	+ # uses the strictest known existing lifetime for every cache layer.
	1328	+ $ttl = (int) METASYNC_OTTO_FALLBACK_CACHE_TTL;
	1329	+
	1330	+ if (!headers_sent()) {
	1331	+ foreach (metasync_otto_fallback_cache_headers($ttl, $reason) as $name => $value) {
	1332	+ header($name . ': ' . $value);
	1333	+ }
	1334	+ }
	1335	+
	1336	+ # Reaches page-cache plugins, which do not see the headers above.
	1337	+ if (!defined('DONOTCACHEPAGE')) {
	1338	+ define('DONOTCACHEPAGE', true);
	1339	+ }
	1340	+}
	1341	+
	1342	+/**
	1343	+ * Report an OTTO render failure to the local error log and to Sentry.
	1344	+ *
	1345	+ * These failures were invisible: the render path swallowed them and returned the
	1346	+ * original HTML, so nobody knew how often they fired. Reporting them is what makes
	1347	+ * a persistently-failing URL discoverable instead of silently costing origin work.
	1348	+ *
	1349	+ * Two constraints shape this:
	1350	+ *
	1351	+ * 1. It runs on the visitor render path, so it must not add latency. The Sentry
	1352	+ * report is therefore queued here and sent by
	1353	+ * metasync_otto_flush_render_failure_reports() after the response has been
	1354	+ * handed to the visitor, so the cost is invisible rather than merely small.
	1355	+ *
	1356	+ * Two alternatives were tried and rejected. Sending inline with a blocking
	1357	+ * transport — cURL with a 5s connect and 5s read timeout — could add ten
	1358	+ * seconds to a page load, precisely when the request is already degraded.
	1359	+ * Sending inline non-blocking is far cheaper but still not free: opening the
	1360	+ * connection and writing the body costs the visitor something.
	1361	+ *
	1362	+ * A WP-Cron single event was the other obvious option and is worse than both:
	1363	+ * every scheduled event is stored in the `cron` option, which is autoloaded on
	1364	+ * every request site-wide, and WP's own 10-minute duplicate suppression keys on
	1365	+ * the serialized args — which carry per-request byte counts here, so it would
	1366	+ * never engage. A site with many failing URLs would grow that option
	1367	+ * unboundedly and pay for it on every page load, which is the cost the deferral
	1368	+ * was meant to avoid in the first place.
	1369	+ *
	1370	+ * The local log stays synchronous. It is cheaper than Sentry but not free —
	1371	+ * Metasync_Error_Logger::log() rewrites a summary option on each call — which is
	1372	+ * the second reason for the throttle below.
	1373	+ *
	1374	+ * 2. A failing page can be requested thousands of times an hour, so reports are
	1375	+ * throttled per reason+URL.
	1376	+ *
	1377	+ * @param string $reason Machine-readable failure reason (e.g. 'RENDER_EXCEPTION').
	1378	+ * @param string $message Human-readable description.
	1379	+ * @param array $context Extra key/value context.
	1380	+ * @param string $level Sentry level: 'error' or 'warning'.
	1381	+ */
	1382	+function metasync_otto_report_render_failure($reason, $message, $context = [], $level = 'error') {
	1383	+ $url = '';
	1384	+ if (!empty($_SERVER['HTTP_HOST']) && !empty($_SERVER['REQUEST_URI'])) {
	1385	+ $url = (is_ssl() ? 'https://' : 'http://')
	1386	+ . $_SERVER['HTTP_HOST']
	1387	+ . strtok(sanitize_text_field(wp_unslash($_SERVER['REQUEST_URI'])), '?');
	1388	+ }
	1389	+
	1390	+ # Throttle: at most one report per reason+URL per 5 minutes. Requests without a
	1391	+ # resolvable URL (CLI, odd SAPIs) collapse onto one key, which is intended —
	1392	+ # they cannot be attributed to a page anyway.
	1393	+ $throttle_key = 'metasync_otto_fail_' . md5($reason . '|' . $url);
	1394	+ if (get_transient($throttle_key)) {
	1395	+ return;
	1396	+ }
	1397	+ set_transient($throttle_key, 1, 5 * MINUTE_IN_SECONDS);
	1398	+
	1399	+ $context = array_merge([
	1400	+ 'reason' => $reason,
	1401	+ 'url' => $url,
	1402	+ ], $context);
	1403	+
	1404	+ # Local structured log. Stays on the customer's site, so it is useful to
	1405	+ # support but not visible to us — hence the Sentry dispatch below.
	1406	+ if (class_exists('Metasync_Error_Logger')) {
	1407	+ Metasync_Error_Logger::log(
	1408	+ Metasync_Error_Logger::CATEGORY_OTTO_RENDER,
	1409	+ $level === 'warning'
	1410	+ ? Metasync_Error_Logger::SEVERITY_WARNING
	1411	+ : Metasync_Error_Logger::SEVERITY_ERROR,
	1412	+ $message,
	1413	+ $context
	1414	+ );
	1415	+ }
	1416	+
	1417	+ # Sentry. Queued rather than sent here: even a non-blocking dispatch costs the
	1418	+ # visitor something (opening the connection and writing the body), and this runs
	1419	+ # on a request that is already degraded. The queue is flushed after the response
	1420	+ # has been handed to the visitor — see metasync_otto_flush_render_failure_reports().
	1421	+ if (!function_exists('metasync_sentry_capture_message_nonblocking')) {
	1422	+ return;
	1423	+ }
	1424	+
	1425	+ $GLOBALS['metasync_otto_pending_failure_reports'][] = [
	1426	+ 'message' => $message,
	1427	+ 'level' => $level,
	1428	+ 'context' => $context,
	1429	+ ];
	1430	+
	1431	+ if (empty($GLOBALS['metasync_otto_failure_flush_hooked'])) {
	1432	+ $GLOBALS['metasync_otto_failure_flush_hooked'] = true;
	1433	+ # PHP_INT_MAX so this runs after everything else registered on shutdown —
	1434	+ # WordPress flushes its own output buffers and closes the object cache at the
	1435	+ # default priority, and this handler ends output for the request.
	1436	+ add_action('shutdown', 'metasync_otto_flush_render_failure_reports', PHP_INT_MAX);
	1437	+ }
	1438	+}
	1439	+
	1440	+/**
	1441	+ * Send any queued render-failure reports after the response has been delivered.
	1442	+ *
	1443	+ * fastcgi_finish_request() (PHP-FPM) and litespeed_finish_request() (LiteSpeed)
	1444	+ * flush the response and close the connection to the visitor while leaving PHP
	1445	+ * running. Calling one of them here means the page is already on its way before
	1446	+ * any telemetry work starts, so the reporting cost is invisible to the visitor
	1447	+ * rather than merely small.
	1448	+ *
	1449	+ * Neither exists under mod_php or the CLI SAPI. There the reports are still sent
	1450	+ * at the very end of the request, after all page output has been generated and
	1451	+ * flushed — the same ordering, just without the connection being closed first.
	1452	+ *
	1453	+ * The dispatch itself stays non-blocking. Now that the visitor is served, the only
	1454	+ * remaining cost is PHP worker occupancy, and a blocking transport with a 5s
	1455	+ * connect and 5s read timeout could tie up a worker for ten seconds per report on
	1456	+ * a site that is failing often.
	1457	+ */
	1458	+function metasync_otto_flush_render_failure_reports() {
	1459	+ $reports = $GLOBALS['metasync_otto_pending_failure_reports'] ?? [];
	1460	+ $GLOBALS['metasync_otto_pending_failure_reports'] = [];
	1461	+
	1462	+ if (empty($reports)) {
	1463	+ return;
	1464	+ }
	1465	+
	1466	+ if (function_exists('fastcgi_finish_request')) {
	1467	+ fastcgi_finish_request();
	1468	+ } elseif (function_exists('litespeed_finish_request')) {
	1469	+ litespeed_finish_request();
	1470	+ }
	1471	+
	1472	+ if (!function_exists('metasync_sentry_capture_message_nonblocking')) {
	1473	+ return;
	1474	+ }
	1475	+
	1476	+ foreach ($reports as $report) {
	1477	+ try {
	1478	+ metasync_sentry_capture_message_nonblocking(
	1479	+ $report['message'],
	1480	+ $report['level'],
	1481	+ $report['context']
	1482	+ );
	1483	+ } catch (Exception $e) {
	1484	+ # Telemetry must never be able to break a request.
	1485	+ } catch (Error $e) {
	1486	+ # Same.
	1487	+ }
	1488	+ }
	1489	+}
	1490	+
	1491	+/**
	1492	+ * Whether MetaSync is still emitting Open Graph / Twitter tags.
	1493	+ *
	1494	+ * The Social Media & Open Graph switch stops MetaSync writing those tags on
	1495	+ * every render path, so the suppression below has to stand down with it —
	1496	+ * removing another plugin's og:/twitter: tag while contributing none of our own
	1497	+ * would leave the page with no social tags at all.
	1498	+ *
	1499	+ * The class_exists() guard keeps this usable if the flags file has not loaded
	1500	+ * yet; treating that window as "enabled" preserves the previous behaviour.
	1501	+ *
	1502	+ * @return bool
	1503	+ */
	1504	+function metasync_otto_social_output_enabled() {
	1505	+ if (!class_exists('Metasync_Feature_Flags')) {
	1506	+ return true;
	1507	+ }
	1508	+
	1509	+ return Metasync_Feature_Flags::is_enabled(Metasync_Feature_Flags::SOCIAL_OG);
	1510	+}
	1511	+
	1512	+/**
779	1513	* Block SEO plugins conditionally based on what Otto is providing
780	1514	* Only blocks title if Otto has title, only blocks description if Otto has description
781	1515	* This prevents duplicate SEO tags while allowing fallback to SEO plugins when Otto has no data.
782	1516	* Supports Yoast SEO, Rank Math, and AIOSEO (free + pro).
		@@ -816,11 +1550,19 @@
816	1550
817	1551
818	1552	# Remove description presenters only when OTTO has a description
819	1553	if ($block_description) {
	1554	+ # The plain meta description belongs to the SEO title/description
	1555	+ # feature, which has its own setting.
820	1556	$presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Meta_Description_Presenter';
821		- $presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Open_Graph\Description_Presenter';
822		- $presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Twitter\Description_Presenter';
	1557	+
	1558	+ # The OG/Twitter descriptions are social tags. With that feature
	1559	+ # switched off MetaSync writes none of its own, so Yoast's must be
	1560	+ # left alone rather than removed with nothing to replace them.
	1561	+ if (metasync_otto_social_output_enabled()) {
	1562	+ $presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Open_Graph\Description_Presenter';
	1563	+ $presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Twitter\Description_Presenter';
	1564	+ }
823	1565	}
824	1566
825	1567	foreach ($presenters as $key => $presenter) {
826	1568	// Safely get class name, suppressing autoload errors
		@@ -838,11 +1580,13 @@
838	1580	# Disable Rank Math
839	1581	if (is_plugin_active('seo-by-rank-math/rank-math.php') ||
840	1582	is_plugin_active('seo-by-rankmath/rank-math.php')) {
841	1583
842		- if ($block_title) {
843		- add_filter('rank_math/frontend/title', '__return_empty_string', 999);
844		- }
	1584	+ # TITLE: Never blank Rank Math's title output during SSR fetch.
	1585	+ # rank_math/frontend/title also gates Rank Math's og:title/twitter:title
	1586	+ # output, so forcing it empty blanks those social tags along with the
	1587	+ # document title. Let Rank Math render its own title/og/twitter tags;
	1588	+ # OTTO's replace_title() and the dedup/precedence pass clean up duplicates.
845	1589
846	1590	if ($block_description) {
847	1591	add_filter('rank_math/frontend/description', '__return_false', 999);
848	1592	add_filter('rank_math/frontend/show_keywords', '__return_false', 999);
		@@ -853,17 +1597,22 @@
853	1597	if (is_plugin_active('all-in-one-seo-pack/all_in_one_seo_pack.php') ||
854	1598	is_plugin_active('all-in-one-seo-pack-pro/all_in_one_seo_pack.php')) {
855	1599
856	1600	if ($block_title) {
	1601	+ # The document title belongs to the SEO title feature, which has its
	1602	+ # own setting; only the og:/twitter: unsets below are social.
857	1603	add_filter('aioseo_title', '__return_empty_string', 999);
858		- add_filter('aioseo_facebook_tags', function($meta) {
859		- if (is_array($meta)) { unset($meta['og:title']); }
860		- return $meta;
861		- }, 999);
862		- add_filter('aioseo_twitter_tags', function($meta) {
863		- if (is_array($meta)) { unset($meta['twitter:title']); }
864		- return $meta;
865		- }, 999);
	1604	+
	1605	+ if (metasync_otto_social_output_enabled()) {
	1606	+ add_filter('aioseo_facebook_tags', function($meta) {
	1607	+ if (is_array($meta)) { unset($meta['og:title']); }
	1608	+ return $meta;
	1609	+ }, 999);
	1610	+ add_filter('aioseo_twitter_tags', function($meta) {
	1611	+ if (is_array($meta)) { unset($meta['twitter:title']); }
	1612	+ return $meta;
	1613	+ }, 999);
	1614	+ }
866	1615	}
867	1616
868	1617	if ($block_description) {
869	1618	# Use granular tag list when available to only block what OTTO provides
		@@ -868,10 +1617,12 @@
868	1617	if ($block_description) {
869	1618	# Use granular tag list when available to only block what OTTO provides
870	1619	$tags = !empty($description_tags) ? $description_tags : [];
871	1620	$block_standard = empty($tags) || in_array('meta[name=description]', $tags);
872		- $block_og_desc = empty($tags) || in_array('meta[property=og:description]', $tags);
873		- $block_tw_desc = empty($tags) || in_array('meta[name=twitter:description]', $tags);
	1621	+ # Social tags only get blocked while MetaSync is still emitting them.
	1622	+ $social_enabled = metasync_otto_social_output_enabled();
	1623	+ $block_og_desc = $social_enabled && (empty($tags) || in_array('meta[property=og:description]', $tags));
	1624	+ $block_tw_desc = $social_enabled && (empty($tags) || in_array('meta[name=twitter:description]', $tags));
874	1625
875	1626	if ($block_standard) {
876	1627	add_filter('aioseo_description', '__return_empty_string', 999);
877	1628	}
		@@ -939,13 +1690,25 @@
939	1690	set_transient($cache_key, 'no', HOUR_IN_SECONDS);
940	1691	return;
941	1692	}
942	1693
943		- $body = wp_remote_retrieve_body($page_data);
944		- $dom = new HtmlDocument($body);
945		- $script = $dom->find('script#sa-dynamic-optimization', 0);
	1694	+ $body = wp_remote_retrieve_body($page_data);
946	1695
947		- if ($script AND !empty($script->getAttribute('data-uuid'))) {
	1696	+ # Detect the OTTO script via a lightweight regex instead of parsing
	1697	+ # the entire fetched page into a SimpleHtmlDom tree (which exhausted memory
	1698	+ # on large pages — this check runs on admin page loads). We only need to
	1699	+ # confirm a <script id="sa-dynamic-optimization" ... data-uuid="..."> exists.
	1700	+ $found = preg_match(
	1701	+ '/<script\b[^>]*\bid=["\']sa-dynamic-optimization["\'][^>]*\bdata-uuid=["\'][^"\']+["\']/i',
	1702	+ (string) $body
	1703	+ )
	1704	+ || preg_match(
	1705	+ # Attribute order may vary (data-uuid before id)
	1706	+ '/<script\b[^>]*\bdata-uuid=["\'][^"\']+["\'][^>]*\bid=["\']sa-dynamic-optimization["\']/i',
	1707	+ (string) $body
	1708	+ );
	1709	+
	1710	+ if ($found) {
948	1711	set_transient($cache_key, 'yes', 12 * HOUR_IN_SECONDS);
949	1712	return;
950	1713	}
951	1714
		@@ -957,8 +1720,17 @@
957	1720
958	1721	# Handle AJAX Clear Cache request
959	1722	# NOTE: Cache system removed - this is now a no-op
960	1723	function metasync_clear_otto_cache_handler() {
	1724	+ # State-changing admin action: verify the nonce sent by admin/js
	1725	+ # (clear_otto_caches) and confirm plugin access before clearing the
	1726	+ # detection transient or queueing the loopback re-check.
	1727	+ check_ajax_referer('metasync_clear_otto_cache', 'nonce');
	1728	+
	1729	+ if (!Metasync::current_user_has_plugin_access()) {
	1730	+ wp_send_json_error(['message' => 'Not authorized']);
	1731	+ }
	1732	+
961	1733	if (!empty($_GET['clear_otto_cache'])) {
962	1734	delete_transient('metasync_otto_js_detected');
963	1735	# Re-run the JS detection in the background so the notice refreshes.
964	1736	metasync_schedule_otto_js_check();
		@@ -1002,15 +1774,21 @@
1002	1774	add_action('wp', 'metasync_start_otto');
1003	1775
1004	1776	# ENHANCED OTTO SEO INTEGRATION
1005	1777	# Register async SEO processing hook
1006		- add_action('metasync_process_seo_job', 'metasync_process_otto_seo_data', 10, 3);
	1778	+ add_action('metasync_process_seo_job', 'metasync_process_otto_seo_data', 10, 4);
1007	1779	add_action('metasync_process_otto_crawl_url_job', 'metasync_handle_otto_crawl_url_job', 10, 2);
1008		- add_action('metasync_process_otto_batch_cache_job', 'metasync_handle_otto_batch_cache_job');
	1780	+ add_action('metasync_process_otto_batch_cache_job', 'metasync_handle_otto_batch_cache_job', 10, 2);
	1781	+ add_action('metasync_otto_pending_drainer', 'metasync_handle_otto_pending_drainer');
1009	1782
1010	1783	# Process OTTO SEO data and update WordPress meta fields for SEO plugins
1011	1784	# This function now runs asynchronously via WordPress cron system
1012	1785	#
	1786	+ # Returns one of the Metasync_Otto_Job_Status::OUTCOME_* strings so callers
	1787	+ # can distinguish success, confirmed no-change, retryable failure, permanent
	1788	+ # failure, and CPU deferral — a bare boolean cannot drive correct retry and
	1789	+ # purge decisions.
	1790	+ #
1013	1791	# @param string $route Fully-qualified URL to process.
1014	1792	# @param bool $allow_defer When false, skip CPU-deferral rescheduling (used when
1015	1793	# called synchronously from crawl_url_job whose own retry
1016	1794	# mechanism already handles failures).
		@@ -1015,10 +1793,14 @@
1015	1793	# called synchronously from crawl_url_job whose own retry
1016	1794	# mechanism already handles failures).
1017	1795	# @param int $deferral_count How many times this job has already been deferred for
1018	1796	# CPU load. Prevents infinite reschedule loops.
	1797	+ # @param array|null $prefetched_seo_data Suggestions already fetched by the transient
	1798	+ # warm step. When an array it is used as-is so the job
	1799	+ # makes exactly ONE API call per URL; the endpoint is
	1800	+ # only fetched here when no prefetch was provided.
1019	1801
1020		-function metasync_process_otto_seo_data($route, $allow_defer = true, $deferral_count = 0) {
	1802	+function metasync_process_otto_seo_data($route, $allow_defer = true, $deferral_count = 0, $prefetched_seo_data = null) {
1021	1803	# Maximum number of times a job can be deferred before it is dropped.
1022	1804	$max_deferrals = defined('METASYNC_SEO_JOB_MAX_DEFERRALS') ? METASYNC_SEO_JOB_MAX_DEFERRALS : 5;
1023	1805	$lock_key = null;
1024	1806	$lock_acquired = false;
		@@ -1025,9 +1807,9 @@
1025	1807
1026	1808	try {
1027	1809	# Validate input
1028	1810	if (empty($route) || !is_string($route)) {
1029		- return false;
	1811	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1030	1812	}
1031	1813
1032	1814	# CPU load check — defer if server is under load.
1033	1815	# Only reschedule when called from the cron hook (allow_defer=true) and
		@@ -1034,24 +1816,28 @@
1034	1816	# we haven't exceeded the maximum deferral count.
1035	1817	# Checked BEFORE acquiring the lock so that deferred jobs don't
1036	1818	# acquire-then-immediately-release it (which defeats concurrency protection).
1037	1819	if (!Metasync_CPU_Monitor::is_load_safe()) {
1038		- if ($allow_defer && $deferral_count < $max_deferrals) {
1039		- wp_schedule_single_event(
1040		- time() + 60,
1041		- 'metasync_process_seo_job',
1042		- array($route, true, $deferral_count + 1)
1043		- );
1044		- return false;
	1820	+ if ($allow_defer) {
	1821	+ if ($deferral_count < $max_deferrals) {
	1822	+ metasync_otto_schedule_single_event(
	1823	+ time() + 60,
	1824	+ 'metasync_process_seo_job',
	1825	+ array($route, true, $deferral_count + 1)
	1826	+ );
	1827	+ return Metasync_Otto_Job_Status::OUTCOME_DEFERRED;
	1828	+ }
	1829	+ # deferral budget exhausted — RUN the job late instead
	1830	+ # of dropping it. The write is idempotent and bounded, and a
	1831	+ # late sync is strictly better than SEO data that never
	1832	+ # arrives. No new cron events are created, so the
	1833	+ # anti-pile-up guarantee is preserved. Fall through.
	1834	+ } else {
	1835	+ # allow_defer=false (sync path): report retryable without creating
	1836	+ # cron events — the crawl-url job's retry mechanism (or OTTO's
	1837	+ # next crawl) re-triggers this URL.
	1838	+ return Metasync_Otto_Job_Status::OUTCOME_RETRYABLE;
1045	1839	}
1046		- # Max deferrals reached or not allowed to defer — drop the job to
1047		- # prevent unbounded cron accumulation that causes server overload.
1048		- if ($deferral_count >= $max_deferrals) {
1049		- metasync_record_failed_action('metasync_process_seo_job');
1050		- return false;
1051		- }
1052		- # allow_defer=false (sync path): just return false, parent handles retry.
1053		- return false;
1054	1840	}
1055	1841
1056	1842	# Concurrency lock: prevent multiple SEO jobs from running simultaneously.
1057	1843	# Uses a per-URL transient lock with a 120s TTL as a safety net (the lock is
		@@ -1059,15 +1845,15 @@
1059	1845	# processing this URL — reschedule once with a short delay instead of stacking.
1060	1846	$lock_key = 'metasync_seo_lock_' . md5($route);
1061	1847	if (get_transient($lock_key) !== false) {
1062	1848	if ($allow_defer && $deferral_count < $max_deferrals) {
1063		- wp_schedule_single_event(
	1849	+ metasync_otto_schedule_single_event(
1064	1850	time() + 30,
1065	1851	'metasync_process_seo_job',
1066	1852	array($route, true, $deferral_count + 1)
1067	1853	);
1068	1854	}
1069		- return false;
	1855	+ return Metasync_Otto_Job_Status::OUTCOME_RETRYABLE;
1070	1856	}
1071	1857	set_transient($lock_key, true, 120);
1072	1858	$lock_acquired = true;
1073	1859
		@@ -1076,16 +1862,15 @@
1076	1862
1077	1863	# Skip excluded URLs - don't process SEO data for them
1078	1864	if (metasync_is_otto_url_excluded($route)) {
1079	1865	//error_log('MetaSync OTTO: Skipping SEO processing for excluded URL: ' . $route);
1080		- return false;
	1866	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1081	1867	}
1082	1868
1083	1869	# Pre-flight 404 check: exclude URLs that would return 404 before making API call
1084	1870	if (!metasync_otto_is_url_available($route)) {
1085		- error_log("MetaSync OTTO: Skipping SEO processing for URL that would return 404: {$route}");
1086	1871	metasync_otto_auto_exclude_404_url($route);
1087		- return false;
	1872	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1088	1873	}
1089	1874
1090	1875	# Get OTTO UUID from settings
1091	1876	# OPTIMIZED: Use cached options
		@@ -1091,37 +1876,37 @@
1091	1876	# OPTIMIZED: Use cached options
1092	1877	$otto_uuid = Metasync_Otto_Config::get_otto_uuid();
1093	1878
1094	1879	if (empty($otto_uuid)) {
1095		- return false;
	1880	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1096	1881	}
1097	1882
1098	1883	# Meta descriptions are always enabled by default - no check needed
1099	1884
1100		- # Fetch SEO data from OTTO API
1101		- $seo_data = metasync_fetch_otto_seo_data($route, $otto_uuid);
1102		-
1103		- if (!$seo_data) {
1104		- metasync_record_failed_action( 'metasync_process_seo_job' );
1105		- return false;
	1885	+ # Obtain the OTTO suggestions for this URL. When the caller already
	1886	+ # warmed the transient cache (crawl-url job), those suggestions are
	1887	+ # handed in and NO second API call is made — the endpoint, rate
	1888	+ # limiter, and breaker are only hit once per URL.
	1889	+ # An empty array is a legitimate answer ("OTTO holds nothing for this
	1890	+ # URL") and must not be treated as a fetch failure; only a strict
	1891	+ # false from the fetcher is a failure.
	1892	+ if (is_array($prefetched_seo_data)) {
	1893	+ $seo_data = $prefetched_seo_data;
	1894	+ } else {
	1895	+ $failure_class = null;
	1896	+ $seo_data = metasync_fetch_otto_seo_data($route, $otto_uuid, $failure_class);
	1897	+ if ($seo_data === false) {
	1898	+ if ($failure_class === Metasync_Otto_Job_Status::OUTCOME_PERMANENT) {
	1899	+ # Auth/input-level rejection from the API — retrying the
	1900	+ # same request cannot help.
	1901	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
	1902	+ }
	1903	+ # Timeout / 5xx / 429 / empty body / bad JSON: retryable.
	1904	+ metasync_record_failed_action( 'metasync_process_seo_job' );
	1905	+ return Metasync_Otto_Job_Status::OUTCOME_RETRYABLE;
	1906	+ }
1106	1907	}
1107	1908
1108		- # Mark this URL as crawled by OTTO for SSR
1109		- # Extract domain and path from route
1110		- $parsed_url = parse_url($route);
1111		- $domain_with_scheme = ($parsed_url['scheme'] ?? 'https') . '://' . ($parsed_url['host'] ?? '');
1112		- $url_path = ($parsed_url['path'] ?? '/');
1113		-
1114		- # Create crawl data structure
1115		- $crawl_data = array(
1116		- 'domain' => $domain_with_scheme,
1117		- 'urls' => array($url_path)
1118		- );
1119		-
1120		- # Load Otto pixel class and save crawl data
1121		- $otto_pixel = new Metasync_otto_pixel($otto_uuid);
1122		- $otto_pixel->save_crawl_data($crawl_data);
1123		-
1124	1909	# Get WordPress post ID from URL
1125	1910	$post_id = url_to_postid($route);
1126	1911
1127	1912	# Special handling for WooCommerce shop page (url_to_postid doesn't work for it)
		@@ -1141,9 +1926,9 @@
1141	1926
1142	1927	# Try to find WooCommerce product by URL if url_to_postid failed
1143	1928	if ((!$post_id || $post_id <= 0) && strpos($route, '/product/') !== false && function_exists('wc_get_products')) {
1144	1929	# Extract product slug from URL
1145		- $product_slug = basename(parse_url($route, PHP_URL_PATH));
	1930	+ $product_slug = basename(wp_parse_url($route, PHP_URL_PATH));
1146	1931
1147	1932	# Try to get product by slug
1148	1933	$products = wc_get_products(array(
1149	1934	'name' => $product_slug,
		@@ -1174,17 +1959,16 @@
1174	1959	if (!$post_id || $post_id <= 0) {
1175	1960	# Check if this is a category page
1176	1961	if (strpos($route, '/category/') !== false) {
1177	1962	# Extract category slug from URL
1178		- $category_slug = basename(parse_url($route, PHP_URL_PATH));
	1963	+ $category_slug = basename(wp_parse_url($route, PHP_URL_PATH));
1179	1964	$category = get_category_by_slug($category_slug);
1180	1965
1181	1966	if ($category) {
1182	1967	# Check if category would return 404 before applying OTTO changes
1183	1968	if (metasync_would_term_return_404($category->term_id, 'category', $route)) {
1184		- error_log("MetaSync OTTO: Skipping SEO processing for category that would return 404: {$route} (Category ID: {$category->term_id})");
1185	1969	metasync_otto_auto_exclude_404_url($route);
1186		- return false;
	1970	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1187	1971	}
1188	1972
1189	1973	# Update comprehensive category SEO meta fields
1190	1974	$update_result = metasync_update_comprehensive_category_seo_fields($category->term_id, $seo_data);
		@@ -1247,8 +2031,12 @@
1247	2031	$title = 'Category Structured Data Update';
1248	2032	break;
1249	2033	}
1250	2034
	2035	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2036	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2037	+ }
	2038	+
1251	2039	if (!empty($title)) {
1252	2040	metasync_log_sync_history([
1253	2041	'title' => $title,
1254	2042	'source' => 'OTTO SEO',
		@@ -1264,12 +2052,12 @@
1264	2052	]);
1265	2053	}
1266	2054	}
1267	2055
1268		- return true;
	2056	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1269	2057	}
1270	2058
1271		- return false;
	2059	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1272	2060	}
1273	2061	}
1274	2062
1275	2063	# Check if this is a WooCommerce product category
		@@ -1274,17 +2062,16 @@
1274	2062
1275	2063	# Check if this is a WooCommerce product category
1276	2064	if (strpos($route, '/product-category/') !== false) {
1277	2065	# Extract product category slug from URL
1278		- $category_slug = basename(parse_url($route, PHP_URL_PATH));
	2066	+ $category_slug = basename(wp_parse_url($route, PHP_URL_PATH));
1279	2067	$term = get_term_by('slug', $category_slug, 'product_cat');
1280	2068
1281	2069	if ($term && !is_wp_error($term)) {
1282	2070	# Check if product category would return 404 before applying OTTO changes
1283	2071	if (metasync_would_term_return_404($term->term_id, 'product_cat', $route)) {
1284		- error_log("MetaSync OTTO: Skipping SEO processing for product category that would return 404: {$route} (Term ID: {$term->term_id})");
1285	2072	metasync_otto_auto_exclude_404_url($route);
1286		- return false;
	2073	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1287	2074	}
1288	2075
1289	2076	# Update comprehensive taxonomy SEO meta fields
1290	2077	$update_result = metasync_update_comprehensive_taxonomy_seo_fields($term->term_id, 'product_cat', $seo_data);
		@@ -1347,8 +2134,12 @@
1347	2134	$title = 'Product Category Structured Data Update';
1348	2135	break;
1349	2136	}
1350	2137
	2138	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2139	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2140	+ }
	2141	+
1351	2142	if (!empty($title)) {
1352	2143	metasync_log_sync_history([
1353	2144	'title' => $title,
1354	2145	'source' => 'OTTO SEO',
		@@ -1365,12 +2156,12 @@
1365	2156	]);
1366	2157	}
1367	2158	}
1368	2159
1369		- return true;
	2160	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1370	2161	}
1371	2162
1372		- return false;
	2163	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1373	2164	}
1374	2165	}
1375	2166
1376	2167	# Check if this is the home page (landing page)
		@@ -1391,11 +2182,10 @@
1391	2182
1392	2183	if ($home_page) {
1393	2184	# Check if home page would return 404 before applying OTTO changes
1394	2185	if (metasync_would_page_return_404($home_page->ID, $route)) {
1395		- error_log("MetaSync OTTO: Skipping SEO processing for home page that would return 404: {$route} (Post ID: {$home_page->ID}, Status: {$home_page->post_status})");
1396	2186	metasync_otto_auto_exclude_404_url($route);
1397		- return false;
	2187	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1398	2188	}
1399	2189
1400	2190	# Update comprehensive home page SEO meta fields
1401	2191	$update_result = metasync_update_comprehensive_seo_fields($home_page->ID, $seo_data);
		@@ -1461,8 +2251,12 @@
1461	2251	$title = 'Home Page Structured Data Update';
1462	2252	break;
1463	2253	}
1464	2254
	2255	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2256	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2257	+ }
	2258	+
1465	2259	if (!empty($title)) {
1466	2260	metasync_log_sync_history([
1467	2261	'title' => $title,
1468	2262	'source' => 'OTTO SEO',
		@@ -1477,12 +2271,12 @@
1477	2271	]);
1478	2272	}
1479	2273	}
1480	2274
1481		- return true;
	2275	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1482	2276	}
1483	2277
1484		- return false;
	2278	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1485	2279	}
1486	2280	}
1487	2281
1488	2282	# Check if this is the blog/posts page (page_for_posts)
		@@ -1493,11 +2287,10 @@
1493	2287
1494	2288	if ($posts_page && $route_clean === $posts_page_url) {
1495	2289	# Check if blog page would return 404
1496	2290	if (metasync_would_page_return_404($posts_page->ID, $route)) {
1497		- error_log("MetaSync OTTO: Skipping SEO processing for blog page that would return 404: {$route} (Post ID: {$posts_page->ID}, Status: {$posts_page->post_status})");
1498	2291	metasync_otto_auto_exclude_404_url($route);
1499		- return false;
	2292	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1500	2293	}
1501	2294
1502	2295	# Update comprehensive blog page SEO meta fields
1503	2296	$update_result = metasync_update_comprehensive_seo_fields($posts_page->ID, $seo_data);
		@@ -1563,8 +2356,12 @@
1563	2356	$title = 'Blog Page Structured Data Update';
1564	2357	break;
1565	2358	}
1566	2359
	2360	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2361	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2362	+ }
	2363	+
1567	2364	if (!empty($title)) {
1568	2365	metasync_log_sync_history([
1569	2366	'title' => $title,
1570	2367	'source' => 'OTTO SEO',
		@@ -1579,12 +2376,12 @@
1579	2376	]);
1580	2377	}
1581	2378	}
1582	2379
1583		- return true;
	2380	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1584	2381	}
1585	2382
1586		- return false;
	2383	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1587	2384	}
1588	2385	}
1589	2386
1590	2387	# URL didn't resolve to any supported entity (post, category, home page, blog page)
		@@ -1589,12 +2386,11 @@
1589	2386
1590	2387	# URL didn't resolve to any supported entity (post, category, home page, blog page)
1591	2388	# Treat as 404 and auto-exclude (e.g. deleted post, non-existent page)
1592	2389	if (!metasync_otto_is_url_available($route)) {
1593		- error_log("MetaSync OTTO: Skipping SEO processing for URL that would return 404 (no matching entity): {$route}");
1594	2390	metasync_otto_auto_exclude_404_url($route);
1595	2391	}
1596		- return false;
	2392	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1597	2393	}
1598	2394
1599	2395	# Verify this is actually a post, page, or WooCommerce product
1600	2396	$post = get_post($post_id);
		@@ -1603,16 +2399,15 @@
1603	2399	$supported_post_types = metasync_get_supported_post_types();
1604	2400
1605	2401	if (!$post || !in_array($post->post_type, $supported_post_types)) {
1606	2402	# Skip unsupported post types
1607		- return false;
	2403	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1608	2404	}
1609	2405
1610	2406	# Check if page would return 404 before applying OTTO changes
1611	2407	if (metasync_would_page_return_404($post_id, $route)) {
1612		- error_log("MetaSync OTTO: Skipping SEO processing for URL that would return 404: {$route} (Post ID: {$post_id}, Status: {$post->post_status})");
1613	2408	metasync_otto_auto_exclude_404_url($route);
1614		- return false;
	2409	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1615	2410	}
1616	2411
1617	2412	# Update comprehensive SEO meta fields
1618	2413	$update_result = metasync_update_comprehensive_seo_fields($post_id, $seo_data);
		@@ -1678,8 +2473,12 @@
1678	2473	$title = 'Structured Data Update';
1679	2474	break;
1680	2475	}
1681	2476
	2477	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2478	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2479	+ }
	2480	+
1682	2481	if (!empty($title)) {
1683	2482	metasync_log_sync_history([
1684	2483	'title' => $title,
1685	2484	'source' => 'OTTO SEO',
		@@ -1694,16 +2493,16 @@
1694	2493	]);
1695	2494	}
1696	2495	}
1697	2496
1698		- return true;
	2497	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1699	2498	}
1700	2499
1701		- return false;
	2500	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1702	2501
1703	2502	} catch (Exception $e) {
1704	2503	metasync_record_failed_action( 'metasync_process_seo_job' );
1705		- return false;
	2504	+ return Metasync_Otto_Job_Status::OUTCOME_RETRYABLE;
1706	2505	} finally {
1707	2506	# Release the concurrency lock only if we actually acquired it.
1708	2507	# Early returns (CPU deferral, lock contention) must NOT delete a lock
1709	2508	# that another process may be holding.
		@@ -1724,11 +2523,10 @@
1724	2523
1725	2524	// Minimal duplicate prevention within short time window
1726	2525	if (!empty($data['title']) && !empty($data['source'])) {
1727	2526	global $wpdb;
1728		- $table = $wpdb->prefix . Metasync_Sync_History_Database::$table_name;
1729		- $recent = $wpdb->get_var($wpdb->prepare(
1730		- "SELECT COUNT(*) FROM `$table` WHERE title = %s AND source = %s AND created_at >= %s",
	2527	+ $recent = $wpdb->get_var($wpdb->prepare( // phpcs:ignore WordPress.DB.DirectDatabaseQuery -- sync-history duplicate guard over the plugin custom table — write-path freshness required
	2528	+ "SELECT COUNT(*) FROM `{$wpdb->prefix}metasync_sync_history` WHERE title = %s AND source = %s AND created_at >= %s",
1731	2529	$data['title'],
1732	2530	$data['source'],
1733	2531	gmdate('Y-m-d H:i:s', time() - 60)
1734	2532	));
		@@ -1739,9 +2537,9 @@
1739	2537
1740	2538	$sync_db->add($data);
1741	2539
1742	2540	} catch (Exception $e) {
1743		- error_log("MetaSync: Failed to log sync history: " . $e->getMessage());
	2541	+ error_log("MetaSync: Failed to log sync history: " . $e->getMessage()); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
1744	2542	}
1745	2543	}
1746	2544
1747	2545	/**
		@@ -1817,9 +2615,9 @@
1817	2615	$db = new Metasync_Redirection_Database();
1818	2616	$redirect = new Metasync_Redirection($db);
1819	2617	return $redirect->resolve_url_to_final_destination($url, 10);
1820	2618	} catch (Exception $e) {
1821		- error_log('MetaSync OTTO: Redirect resolution failed for ' . $url . ' - ' . $e->getMessage());
	2619	+ error_log('MetaSync OTTO: Redirect resolution failed for ' . $url . ' - ' . $e->getMessage()); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
1822	2620	return $url;
1823	2621	}
1824	2622	}
1825	2623
		@@ -1850,9 +2648,9 @@
1850	2648	'status' => 'active',
1851	2649	'auto_excluded' => 1,
1852	2650	]);
1853	2651	} catch (Exception $e) {
1854		- error_log('MetaSync OTTO: Failed to auto-exclude 404 URL: ' . $url . ' - ' . $e->getMessage());
	2652	+ error_log('MetaSync OTTO: Failed to auto-exclude 404 URL: ' . $url . ' - ' . $e->getMessage()); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
1855	2653	return false;
1856	2654	}
1857	2655	}
1858	2656
		@@ -1872,14 +2670,13 @@
1872	2670	try {
1873	2671	require_once plugin_dir_path(__FILE__) . 'class-metasync-otto-excluded-urls-database.php';
1874	2672	$db = new Metasync_Otto_Excluded_URLs_Database();
1875	2673	global $wpdb;
1876		- $table = $wpdb->prefix . Metasync_Otto_Excluded_URLs_Database::$table_name;
1877	2674	# Normalize the same way is_url_excluded() does
1878	2675	$url_normalized = rtrim(trim($url), '/');
1879		- $records = $wpdb->get_results(
	2676	+ $records = $wpdb->get_results( // phpcs:ignore WordPress.DB.DirectDatabaseQuery -- webhook-path cleanup of auto-exclusions over the plugin custom table
1880	2677	$wpdb->prepare(
1881		- "SELECT id FROM `{$table}` WHERE url_pattern = %s AND auto_excluded = 1 AND status = 'active'",
	2678	+ "SELECT id FROM `{$wpdb->prefix}metasync_otto_excluded_urls` WHERE url_pattern = %s AND auto_excluded = 1 AND status = 'active'",
1882	2679	$url_normalized
1883	2680	)
1884	2681	);
1885	2682	if (!empty($records)) {
		@@ -1909,24 +2706,26 @@
1909	2706	}
1910	2707	try {
1911	2708	require_once plugin_dir_path(__FILE__) . 'class-metasync-otto-excluded-urls-database.php';
1912	2709	global $wpdb;
1913		- $table = $wpdb->prefix . Metasync_Otto_Excluded_URLs_Database::$table_name;
1914	2710	$url_normalized = rtrim(trim($url), '/');
1915	2711
1916	2712	$records = get_transient(METASYNC_OTTO_EXCLUDED_TRANSIENT_KEY);
1917	2713
1918	2714	if ($records === false) {
1919		- // phpcs:ignore WordPress.DB.PreparedSQL.NotPrepared -- no user input, table name from $wpdb->prefix
1920		- $records = $wpdb->get_results(
1921		- "SELECT url_pattern, pattern_type FROM `{$table}` WHERE status = 'active' AND (auto_excluded = 0 OR auto_excluded IS NULL) ORDER BY created_at DESC"
	2715	+ $records = $wpdb->get_results( // phpcs:ignore WordPress.DB.DirectDatabaseQuery -- render-path read IS cached via get/set_transient with invalidation on every write — wp_cache_* is not used because the payload must survive object-cache-only setups
	2716	+ "SELECT url_pattern, pattern_type FROM `{$wpdb->prefix}metasync_otto_excluded_urls` WHERE status = 'active' AND (auto_excluded = 0 OR auto_excluded IS NULL) ORDER BY created_at DESC"
1922	2717	);
1923	2718
1924	2719	// Graceful recovery: auto_excluded column missing on pre-v2.7.4 installs.
1925		- // Run ALTER TABLE to add it and treat URL as not excluded so OTTO continues rendering.
	2720	+ // Schema changes belong to the upgrade routines, never the render path:
	2721	+ // degrade to "not excluded" so OTTO keeps rendering, and schedule a
	2722	+ // one-shot background repair that runs the same column migration off
	2723	+ // the visitor request (handled by metasync_repair_otto_excluded_schema).
1926	2724	if ($wpdb->last_error && strpos($wpdb->last_error, 'auto_excluded') !== false) {
1927		- // phpcs:ignore WordPress.DB.PreparedSQL.NotPrepared
1928		- $wpdb->query("ALTER TABLE `{$table}` ADD COLUMN `auto_excluded` TINYINT(1) NOT NULL DEFAULT 0");
	2725	+ if (false === wp_next_scheduled('metasync_repair_otto_excluded_schema')) {
	2726	+ wp_schedule_single_event(time() + MINUTE_IN_SECONDS, 'metasync_repair_otto_excluded_schema');
	2727	+ }
1929	2728	return false;
1930	2729	}
1931	2730
1932	2731	set_transient(METASYNC_OTTO_EXCLUDED_TRANSIENT_KEY, $records ?: [], METASYNC_OTTO_EXCLUDED_TRANSIENT_TTL);
		@@ -1936,27 +2735,16 @@
1936	2735	return false;
1937	2736	}
1938	2737
1939	2738	foreach ($records as $excluded) {
1940		- $pattern = rtrim(trim($excluded->url_pattern), '/');
1941		- $pattern_type = $excluded->pattern_type;
1942		-
1943		- switch ($pattern_type) {
1944		- case 'exact':
1945		- if ($url_normalized === $pattern) {
1946		- return true;
1947		- }
1948		- break;
1949		- case 'contain':
1950		- if (strpos($url_normalized, $pattern) !== false) {
1951		- return true;
1952		- }
1953		- break;
1954		- case 'start':
1955		- if (strpos($url_normalized, $pattern) === 0) {
1956		- return true;
1957		- }
1958		- break;
	2739	+ # Delegate to the shared matcher so this gate supports every pattern type the
	2740	+ # add handler accepts (exact|contain|start|end|regex). Keeping a second switch
	2741	+ # here is what let `end` and `regex` exclusions be silently ignored on the
	2742	+ # visitor render path while still applying on the queue/SEO-write path.
	2743	+ # The raw pattern is passed through - the matcher normalizes it per type,
	2744	+ # because a regex must keep the trailing slash that closes its delimiters.
	2745	+ if (Metasync_Otto_Excluded_URLs_Database::pattern_matches($url_normalized, $excluded->url_pattern, $excluded->pattern_type)) {
	2746	+ return true;
1959	2747	}
1960	2748	}
1961	2749	return false;
1962	2750	} catch (Exception $e) {
		@@ -2013,9 +2801,9 @@
2013	2801	}
2014	2802
2015	2803	# WooCommerce product by slug
2016	2804	if ((!$post_id || $post_id <= 0) && strpos($route, '/product/') !== false && function_exists('wc_get_products')) {
2017		- $product_slug = basename(parse_url($route, PHP_URL_PATH));
	2805	+ $product_slug = basename(wp_parse_url($route, PHP_URL_PATH));
2018	2806	$products = wc_get_products(array('name' => $product_slug, 'limit' => 1, 'status' => 'publish'));
2019	2807	if (!empty($products)) {
2020	2808	$post_id = $products[0]->get_id();
2021	2809	} else {
		@@ -2042,9 +2830,9 @@
2042	2830	}
2043	2831
2044	2832	# Category
2045	2833	if (strpos($route, '/category/') !== false) {
2046		- $category_slug = basename(parse_url($route, PHP_URL_PATH));
	2834	+ $category_slug = basename(wp_parse_url($route, PHP_URL_PATH));
2047	2835	$category = get_category_by_slug($category_slug);
2048	2836	if ($category) {
2049	2837	return !metasync_would_term_return_404($category->term_id, 'category', $route);
2050	2838	}
		@@ -2051,9 +2839,9 @@
2051	2839	}
2052	2840
2053	2841	# WooCommerce product category
2054	2842	if (strpos($route, '/product-category/') !== false) {
2055		- $category_slug = basename(parse_url($route, PHP_URL_PATH));
	2843	+ $category_slug = basename(wp_parse_url($route, PHP_URL_PATH));
2056	2844	$term = get_term_by('slug', $category_slug, 'product_cat');
2057	2845	if ($term && !is_wp_error($term)) {
2058	2846	return !metasync_would_term_return_404($term->term_id, 'product_cat', $route);
2059	2847	}
		@@ -2095,9 +2883,9 @@
2095	2883
2096	2884	$removed = 0;
2097	2885	$marked_permanent = 0;
2098	2886	$thirty_days_ago = strtotime('-30 days');
2099		- $next_recheck = date('Y-m-d H:i:s', current_time('timestamp') + 7 * DAY_IN_SECONDS);
	2887	+ $next_recheck = gmdate('Y-m-d H:i:s', current_time('timestamp') + 7 * DAY_IN_SECONDS);
2100	2888
2101	2889	foreach ($records as $record) {
2102	2890	$url = trim($record->url_pattern);
2103	2891	if (empty($url)) {
		@@ -2118,16 +2906,10 @@
2118	2906	}
2119	2907	}
2120	2908	}
2121	2909
2122		- if ($removed > 0) {
2123		- error_log("MetaSync OTTO: Recheck 404 exclusions - removed {$removed} URL(s) that are now available");
2124		- }
2125		- if ($marked_permanent > 0) {
2126		- error_log("MetaSync OTTO: Recheck 404 exclusions - marked {$marked_permanent} URL(s) as permanent (still 404 after 30 days)");
2127		- }
2128	2910	} catch (Exception $e) {
2129		- error_log('MetaSync OTTO: Recheck 404 exclusions failed - ' . $e->getMessage());
	2911	+ error_log('MetaSync OTTO: Recheck 404 exclusions failed - ' . $e->getMessage()); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
2130	2912	}
2131	2913	}
2132	2914
2133	2915	add_action('metasync_otto_recheck_404_exclusions', 'metasync_otto_recheck_404_exclusions');
		@@ -2132,8 +2914,24 @@
2132	2914
2133	2915	add_action('metasync_otto_recheck_404_exclusions', 'metasync_otto_recheck_404_exclusions');
2134	2916
2135	2917	/**
	2918	+ * Background repair of the metasync_otto_excluded_urls schema (cron).
	2919	+ *
	2920	+ * Scheduled by metasync_is_otto_url_manually_excluded() when a render-path
	2921	+ * read fails because the auto_excluded column is missing (pre-2.7.4 install
	2922	+ * whose version migration did not run). Runs the DB layer's self-healing
	2923	+ * column migration off the visitor request instead of ALTERing inline.
	2924	+ */
	2925	+function metasync_repair_otto_excluded_schema()
	2926	+{
	2927	+ require_once plugin_dir_path(__FILE__) . 'class-metasync-otto-excluded-urls-database.php';
	2928	+ Metasync_Otto_Excluded_URLs_Database::ensure_schema();
	2929	+}
	2930	+
	2931	+add_action('metasync_repair_otto_excluded_schema', 'metasync_repair_otto_excluded_schema');
	2932	+
	2933	+/**
2136	2934	* Check if a post/page would return 404 without making HTTP request
2137	2935	* Uses WordPress database checks for fast validation
2138	2936	*
2139	2937	* @param int $post_id WordPress post ID
		@@ -2241,5 +3039,88 @@
2241	3039	# Check if this post has Brizy metadata
2242	3040	if (get_post_meta($post_id, 'brizy_post_uid', true)) {
2243	3041	delete_transient('metasync_has_brizy_posts');
2244	3042	}
2245		-}, 10, 1);
	3043	+}, 10, 1);
	3044	+
	3045	+/**
	3046	+ * Identify infrastructure agents that must never hit the render throttle.
	3047	+ *
	3048	+ * Two classes of agent are covered:
	3049	+ *
	3050	+ * - Page-cache warmers. A host preloader (WP Cloud PageCacheBot, WP Rocket preload,
	3051	+ * LiteSpeed, SG Optimizer, ...) exists purely to populate the page cache. Throttling
	3052	+ * one returns un-OTTO'd HTML *and* defines DONOTCACHEPAGE, so the cache can never be
	3053	+ * filled and every subsequent human visitor pays a full uncached render. Measured on
	3054	+ * the reporting site: desktop cache MISS on 17/17 requests at ~2.8s TTFB, versus
	3055	+ * ~0.56s with OTTO disabled.
	3056	+ *
	3057	+ * - Synthetic performance auditors (Lighthouse / PageSpeed Insights / GTmetrix).
	3058	+ * These are what customers measure with, so they must receive the same fully
	3059	+ * rendered OTTO output a real visitor gets. Currently `/lighthouse/i` and
	3060	+ * `/pagespeed/i` are generic-bot patterns, so a second PSI run inside the 5-minute
	3061	+ * window is served the throttled, un-OTTO'd, uncacheable page.
	3062	+ *
	3063	+ * Matching is done on the raw user-agent, deliberately independent of the bot-name
	3064	+ * categorisation in Metasync_Otto_Bot_Detector.
	3065	+ *
	3066	+ * @param mixed $detection Result of Metasync_Otto_Bot_Detector::detect(), or malformed input.
	3067	+ * @return bool True when this request must bypass the render throttle.
	3068	+ */
	3069	+function metasync_otto_is_unthrottled_infrastructure_agent( $detection = array() ) {
	3070	+
	3071	+ $ua = '';
	3072	+ if ( is_array( $detection ) && ! empty( $detection['user_agent'] ) ) {
	3073	+ $ua = (string) $detection['user_agent'];
	3074	+ } elseif ( isset( $_SERVER['HTTP_USER_AGENT'] ) ) {
	3075	+ $ua = (string) $_SERVER['HTTP_USER_AGENT'];
	3076	+ }
	3077	+
	3078	+ $needles = array(
	3079	+ // Page-cache preloaders.
	3080	+ 'pagecachebot', // WP Cloud / Automattic (PageCacheBotDesktop|Mobile)
	3081	+ 'wp rocket', // WP Rocket preload
	3082	+ 'wprocket',
	3083	+ 'lscache_runner', // LiteSpeed Cache crawler
	3084	+ 'sg-optimizer', // SiteGround
	3085	+ 'sgoptimizer',
	3086	+ 'nitropack',
	3087	+ 'cache-warmer',
	3088	+ 'cachewarmer',
	3089	+ 'cache warmer',
	3090	+ // Synthetic performance auditors.
	3091	+ 'lighthouse',
	3092	+ 'pagespeed',
	3093	+ 'gtmetrix',
	3094	+ 'pingdom',
	3095	+ 'webpagetest',
	3096	+ );
	3097	+
	3098	+ $is_infra = false;
	3099	+
	3100	+ if ( '' !== $ua ) {
	3101	+ $ua_lower = strtolower( $ua );
	3102	+ foreach ( $needles as $needle ) {
	3103	+ if ( false !== strpos( $ua_lower, $needle ) ) {
	3104	+ $is_infra = true;
	3105	+ break;
	3106	+ }
	3107	+ }
	3108	+ }
	3109	+
	3110	+ // WP Cloud's warmer also tags its requests with a query marker.
	3111	+ if ( ! $is_infra && isset( $_GET['x-cache-engine'] ) ) { // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only flag for render throttling
	3112	+ $is_infra = true;
	3113	+ }
	3114	+
	3115	+ /**
	3116	+ * Allow a site or host to declare additional preload agents.
	3117	+ *
	3118	+ * @param bool $is_infra Whether this request bypasses the OTTO render throttle.
	3119	+ * @param string $ua The request user-agent.
	3120	+ */
	3121	+ if ( function_exists( 'apply_filters' ) ) {
	3122	+ $is_infra = (bool) apply_filters( 'metasync_otto_unthrottled_agent', $is_infra, $ua );
	3123	+ }
	3124	+
	3125	+ return $is_infra;
	3126	+}