← All changes | otto/otto_pixel.php +1070 -248 2.6.20 → 2.7.1 View file →

		@@ -16,18 +16,66 @@
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';
25	48	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-render-strategy.php';
26	49	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-config.php';
27	50	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-bot-detector.php';
28	51	require_once plugin_dir_path( __FILE__ ) . '/class-metasync-otto-bot-statistics-database.php';
29	52
	53	+/**
	54	+ * Is $method actually available on $class in this process right now?
	55	+ *
	56	+ * A partially updated install can leave a newer copy of one plugin file beside
	57	+ * an older copy of another — stale opcache bytecode for a single file is enough.
	58	+ * class_exists() is satisfied by the older copy, and calling a method it does
	59	+ * not declare is a fatal. Everything in this file runs on the front end, so that
	60	+ * fatal is a white screen on every page view rather than a degraded feature.
	61	+ *
	62	+ * Defined here rather than pulled from a shared utility class on purpose: a
	63	+ * helper loaded from another file could itself be the stale one.
	64	+ *
	65	+ * $class and $method are parameters rather than literals at the call site so the
	66	+ * check survives static analysis, which would otherwise narrow a literal
	67	+ * method_exists() on a known class to a constant true — the skew this guards
	68	+ * against exists only at runtime.
	69	+ *
	70	+ * @param string $class Class about to be called.
	71	+ * @param string $method Method about to be called on it.
	72	+ * @return bool
	73	+ */
	74	+function metasync_otto_class_provides($class, $method) {
	75	+ return class_exists($class) && method_exists($class, $method);
	76	+}
	77	+
30	78	# OPTIMIZED: get the metasync options (cached in static class)
31	79	$metasync_options = Metasync_Otto_Config::get_options();
32	80
33	81	# OTTO SSR is always enabled by default
		@@ -51,9 +99,9 @@
51	99	# Use endpoint manager to get the correct domain
52	100	$otto_domain = 'sa.searchatlas.com'; # default
53	101	if (class_exists('Metasync_Endpoint_Manager')) {
54	102	$otto_api_domain = Metasync_Endpoint_Manager::get_endpoint('OTTO_API_DOMAIN');
55		- $parsed = parse_url($otto_api_domain);
	103	+ $parsed = wp_parse_url($otto_api_domain);
56	104	if (!empty($parsed['host'])) {
57	105	$otto_domain = $parsed['host'];
58	106	}
59	107	}
		@@ -69,8 +117,49 @@
69	117	echo $otto_tag;
70	118	}, 1); # Priority 1 to output early in head
71	119
72	120	/**
	121	+ * Scheduling seam for the OTTO pipeline.
	122	+ *
	123	+ * wp_schedule_single_event() returns false on failure and the old pipeline
	124	+ * ignored it, so a broken cron table acknowledged URLs that were never
	125	+ * queued. Every schedule in this pipeline goes through this seam so the
	126	+ * return value is honored (and tests can observe scheduling).
	127	+ */
	128	+function metasync_otto_schedule_single_event($timestamp, $hook, $args = array()) {
	129	+ return wp_schedule_single_event($timestamp, $hook, $args) === true;
	130	+}
	131	+
	132	+/**
	133	+ * Dedupe seam: does this URL already have live per-URL work in flight?
	134	+ *
	135	+ * Combines the durable status store with the raw cron table so a re-delivered
	136	+ * webhook cannot stack a second job behind a live one.
	137	+ */
	138	+function metasync_otto_job_already_scheduled($route) {
	139	+ if (class_exists('Metasync_Otto_Job_Status', false)
	140	+ && Metasync_Otto_Job_Status::has_active($route)) {
	141	+ return true;
	142	+ }
	143	+ return wp_next_scheduled('metasync_process_otto_crawl_url_job', array($route)) !== false;
	144	+}
	145	+
	146	+/**
	147	+ * Cron-table dedupe seam: does a per-URL cron event already exist for this URL?
	148	+ *
	149	+ * The pending-queue drainer must dedupe against the scheduler table only.
	150	+ * Every URL it takes off the durable queue carries a fresh webhook
	151	+ * accepted/batch_overflow ledger row; that row exists so a re-delivered
	152	+ * webhook cannot stack a second job behind live work, and consulting it
	153	+ * from the drainer — after take_pending() has already removed the URL
	154	+ * from the queue — silently dropped every overflow URL instead of
	155	+ * draining it. Only a real, still-pending cron event may veto here.
	156	+ */
	157	+function metasync_otto_job_event_pending($route) {
	158	+ return wp_next_scheduled('metasync_process_otto_crawl_url_job', array($route)) !== false;
	159	+}
	160	+
	161	+/**
73	162	* Start end point to handle requests on page updates
74	163	* The Otto Crawler will call this end point once a page is updated
75	164	**/
76	165
		@@ -105,28 +194,25 @@
105	194	$otto_pixel->save_crawl_data($data);
106	195
107	196	# Defer the expensive per-URL work (OTTO API fetch, post meta sync,
108	197	# 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.
	198	+ # strict response-time budget, so the response acknowledges SCHEDULING,
	199	+ # never execution. The per-URL status store records what actually
	200	+ # happened to each URL once the jobs run.
112	201	#
113		- # 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.
	202	+ # Cap the number of per-URL cron jobs scheduled per webhook batch so a
	203	+ # large crawl cannot overwhelm WP-Cron on shared hosts. URLs past the cap
	204	+ # are queued durably and worked off by a self-rescheduling drainer job —
	205	+ # never silently dropped.
118	206	$max_jobs_per_batch = defined('METASYNC_MAX_JOBS_PER_BATCH') ? METASYNC_MAX_JOBS_PER_BATCH : 25;
119	207	$now = time();
120	208	$routes_to_process = array();
121		- $scheduled_count = 0;
	209	+ $accepted = 0;
	210	+ $deferred = 0;
	211	+ $rejected = 0;
	212	+ $scheduled_this_batch = 0;
122	213
123	214	foreach($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	215	# prepare the route
130	216	$route = $data['domain'] . $url;
131	217
132	218	# validate the route
		@@ -145,39 +231,71 @@
145	231	if (metasync_is_otto_url_excluded($route)) {
146	232	continue;
147	233	}
148	234
	235	+ # Every valid URL belongs to the batched cache-purge set even when its
	236	+ # per-URL meta-sync job is deduped or deferred: the crawler hit all of
	237	+ # them, so the caches for all of them must be refreshed.
	238	+ $routes_to_process[] = $route;
	239	+
	240	+ # Duplicate delivery: an active job already owns this URL. Idempotent
	241	+ # no-op — not re-scheduled, not queued, not counted as new work.
	242	+ if (metasync_otto_job_already_scheduled($route)) {
	243	+ continue;
	244	+ }
	245	+
	246	+ if ($scheduled_this_batch >= $max_jobs_per_batch) {
	247	+ # Batch overflow: queue durably for the drainer instead of dropping.
	248	+ if (Metasync_Otto_Job_Status::enqueue_pending($route)) {
	249	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'queue', 0, 'batch_overflow');
	250	+ $deferred++;
	251	+ } else {
	252	+ $rejected++;
	253	+ }
	254	+ continue;
	255	+ }
	256	+
149	257	# Queue the per-URL processing for background execution. The handler
150	258	# (metasync_handle_otto_crawl_url_job) performs transient warming,
151	259	# SEO meta sync, and per-URL host cache purge.
152	260	# Offset each event by $key seconds to avoid wp_schedule_single_event()
153	261	# silently dropping duplicates when timestamp + hook + args collide.
154		- wp_schedule_single_event($now + $key, 'metasync_process_otto_crawl_url_job', array($route));
155		-
156		- $routes_to_process[] = $route;
157		- $scheduled_count++;
	262	+ # A false return means cron refused the event — fall back to the
	263	+ # durable queue rather than acknowledging a URL nobody will process.
	264	+ if (metasync_otto_schedule_single_event($now + $key, 'metasync_process_otto_crawl_url_job', array($route))) {
	265	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'queue', 0, '');
	266	+ $accepted++;
	267	+ $scheduled_this_batch++;
	268	+ } elseif (Metasync_Otto_Job_Status::enqueue_pending($route)) {
	269	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'queue', 0, 'cron_refused');
	270	+ $deferred++;
	271	+ } else {
	272	+ $rejected++;
	273	+ }
158	274	}
159	275
160		- # Schedule a single batch job for cache warming and edge CDN purge.
161		- # These operations benefit from batching: warm_urls() uses concurrent
162		- # fire-and-forget requests, and edge providers (Cloudflare, Fastly)
163		- # support multi-URL purge in a single API call.
	276	+ # Schedule a single batch job for the edge CDN purge.
	277	+ # The batch waits for per-URL stragglers (retries) before purging, so a
	278	+ # retried URL's fresh content is what reaches the edge. Multi-URL purge
	279	+ # in a single API call keeps edge-provider request counts low.
164	280	if (!empty($routes_to_process)) {
165	281	$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));
	282	+ metasync_otto_schedule_single_event($batch_time, 'metasync_process_otto_batch_cache_job', array($routes_to_process));
167	283	}
168	284
169		- # Track OTTO optimization event in GA4 (local, non-blocking)
170		- try {
171		- Metasync_GA4::get_instance()->track_otto_optimization($data);
172		- } catch (Exception $e) {
173		- // Analytics tracking failed, continue
	285	+ # Work off any overflow (or cron-refused URLs) from the durable queue.
	286	+ if ($deferred > 0) {
	287	+ metasync_otto_schedule_single_event($now + 120, 'metasync_otto_pending_drainer', array());
174	288	}
175	289
176		- # Return 200 immediately so the webhook caller does not time out.
	290	+ # Return 200 immediately so the webhook caller does not time out, with an
	291	+ # honest accounting of what happened to the batch.
177	292	return new WP_REST_Response(array(
178	293	'success' => true,
179	294	'message' => 'OTTO crawl notification received',
	295	+ 'accepted' => $accepted,
	296	+ 'deferred' => $deferred,
	297	+ 'rejected' => $rejected,
180	298	), 200);
181	299	}
182	300
183	301	/**
		@@ -182,119 +300,250 @@
182	300
183	301	/**
184	302	* Background handler for a single OTTO crawl-notify URL.
185	303	*
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.
	304	+ * Pipeline per URL:
	305	+ * 1. Warm the OTTO transient cache — this is the ONE live API call the job
	306	+ * makes; the warmed suggestions are handed to the meta sync so the
	307	+ * endpoint is never fetched twice per URL.
	308	+ * 2. Sync SEO post meta (metasync_process_otto_seo_data) and branch on its
	309	+ * outcome: only SUCCESS / NO_CHANGE reach the purge step.
	310	+ * 3. Clear the per-URL host cache and re-warm it.
188	311	*
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).
	312	+ * Retryable outcomes (API timeout / 5xx / 429 / lock contention / CPU deferral)
	313	+ * retry with exponential backoff up to METASYNC_OTTO_JOB_MAX_RETRIES; the
	314	+ * exhaustion event is recorded via metasync_record_failed_action() for
	315	+ * Site Health. Permanent outcomes (not connected, excluded/404 URL, API auth
	316	+ * or input error) never retry and never purge.
192	317	*
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	318	* @param string $route Fully-qualified URL to process.
198	319	* @param int $retry_count Current retry attempt (0 = first run).
199	320	*/
200		-function metasync_handle_otto_crawl_url_job($route, $retry_count = 0) {
	321	+function metasync_handle_otto_crawl_url_job($route = '', $retry_count = 0) {
	322	+ // WP-Cron events persist independently of plugin code. A stale or malformed
	323	+ // event may therefore invoke this callback without its required route.
	324	+ // Docblock type isn't enforced at runtime; a stale cron record can pass a non-string.
	325	+ // @phpstan-ignore-next-line function.alreadyNarrowedType
	326	+ if (!is_string($route) || $route === '') {
	327	+ 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
	328	+ return;
	329	+ }
	330	+
	331	+ $max_retries = defined('METASYNC_OTTO_JOB_MAX_RETRIES') ? METASYNC_OTTO_JOB_MAX_RETRIES : 3;
	332	+
	333	+ # Shared retry/terminal bookkeeping so every failure path stays consistent.
	334	+ $schedule_retry = function ($reason) use ($route, $retry_count, $max_retries) {
	335	+ if ($retry_count < $max_retries) {
	336	+ # Exponential backoff (60s, 120s, 240s) plus jitter so a batch of
	337	+ # failures cannot stampede the API on the same tick.
	338	+ $delay = 60 * pow(2, $retry_count) + wp_rand(0, 15);
	339	+ $scheduled = metasync_otto_schedule_single_event(time() + $delay, 'metasync_process_otto_crawl_url_job', array($route, $retry_count + 1));
	340	+ if (!$scheduled && Metasync_Otto_Job_Status::enqueue_pending($route)) {
	341	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'queue', $retry_count + 1, $reason . '/cron_refused');
	342	+ } else {
	343	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_RETRYING, 'job', $retry_count + 1, $reason);
	344	+ }
	345	+ error_log('MetaSync OTTO: retrying crawl-url job for ' . $route . ' (attempt ' . ($retry_count + 1) . '/' . $max_retries . ') reason=' . $reason . ' in ' . $delay . 's'); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
	346	+ } else {
	347	+ metasync_record_failed_action('metasync_process_otto_crawl_url_job');
	348	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_FAILED, 'job', $retry_count, 'exhausted:' . $reason);
	349	+ 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
	350	+ }
	351	+ };
	352	+
201	353	try {
202		- # Step 1: Warm OTTO transient cache (fetch fresh suggestions from OTTO API into WP transient)
	354	+ # Not connected: retrying cannot help, and there is nothing to purge.
203	355	$otto_uuid = Metasync_Otto_Config::get_otto_uuid();
204		- if (!empty($otto_uuid)) {
205		- $transient_cache = new Metasync_Otto_Transient_Cache($otto_uuid);
206		- $transient_cache->warm_cache($route);
	356	+ # A corrupted option row can hold a non-string UUID; the docblock
	357	+ # contract alone does not enforce that at runtime.
	358	+ # @phpstan-ignore-next-line function.alreadyNarrowedType
	359	+ if (!is_string($otto_uuid) || $otto_uuid === '') {
	360	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_FAILED, 'job', $retry_count, 'not_connected');
	361	+ return;
207	362	}
208	363
	364	+ # Step 1: Warm OTTO transient cache (fetch fresh suggestions from OTTO
	365	+ # API into WP transient — the rate limiter, circuit breaker, and
	366	+ # request lock all live inside this call).
	367	+ $transient_cache = new Metasync_Otto_Transient_Cache($otto_uuid);
	368	+ $warmed = $transient_cache->warm_cache($route);
	369	+ if ($warmed === false) {
	370	+ # A 4xx refusal (bad key, revoked project) is permanent: retrying
	371	+ # cannot change the answer, so fail fast instead of burning the
	372	+ # attempt budget on warm calls that will keep being rejected.
	373	+ if ($transient_cache->last_failure_is_permanent()) {
	374	+ metasync_record_failed_action('metasync_process_otto_crawl_url_job');
	375	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_FAILED, 'job', $retry_count, 'permanent:warm_rejected');
	376	+ 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
	377	+ return;
	378	+ }
	379	+ # Transient API failure (timeout / 5xx / breaker open). No meta
	380	+ # write, no purge — straight to the retry path.
	381	+ $schedule_retry('warm_failed');
	382	+ return;
	383	+ }
	384	+
209	385	# Step 2: Write SEO post meta synchronously BEFORE the cache purge.
210		- # This ensures the DB is fully up-to-date when Kinsta (or any host) re-populates
211		- # the cache on the very next request — eliminating the race condition where the
212		- # 1-second scheduled job hadn't run yet and stale meta got cached.
213		- # allow_defer=false: do NOT reschedule a new metasync_process_seo_job cron
214		- # event from this synchronous path — the crawl_url_job retry mechanism already
215		- # handles failures, and rescheduling here caused an unbounded cron pile-up.
216		- metasync_process_otto_seo_data($route, false);
	386	+ # This ensures the DB is fully up-to-date when the host re-populates
	387	+ # the cache on the very next request. The warmed suggestions are
	388	+ # passed through so the sync makes NO second API call.
	389	+ # allow_defer=false: the crawl_url_job retry mechanism owns failure
	390	+ # handling here; letting the sync self-reschedule caused an
	391	+ # unbounded cron pile-up.
	392	+ # warm_cache() may hand back a non-array truthy on contract drift;
	393	+ # the sync treats that as "no prefetched suggestions".
	394	+ # @phpstan-ignore-next-line function.alreadyNarrowedType
	395	+ $prefetched = is_array($warmed) ? $warmed : null;
	396	+ $outcome = metasync_process_otto_seo_data($route, false, 0, $prefetched);
217	397
218		- # Step 3: Clear the per-URL cache entry
	398	+ if ($outcome === Metasync_Otto_Job_Status::OUTCOME_RETRYABLE
	399	+ || $outcome === Metasync_Otto_Job_Status::OUTCOME_DEFERRED) {
	400	+ $schedule_retry('outcome_' . $outcome);
	401	+ return;
	402	+ }
	403	+
	404	+ if ($outcome !== Metasync_Otto_Job_Status::OUTCOME_SUCCESS
	405	+ && $outcome !== Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE) {
	406	+ # Permanent failure: no retry, no purge — caches must keep serving
	407	+ # the last good content.
	408	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_FAILED, 'meta', $retry_count, 'permanent:' . $outcome);
	409	+ return;
	410	+ }
	411	+
	412	+ # Step 3: meta is current (changed or confirmed unchanged) — clear the
	413	+ # per-URL host cache entry and re-warm it so OTTO-modified output is
	414	+ # what gets stored.
219	415	$otto_pixel = new Metasync_otto_pixel(false);
220	416	$otto_pixel->refresh_cache($route);
221	417	Metasync_Cache_Purge::purge_single_url($route);
	418	+ Metasync_Cache_Purge::warm_urls(array($route));
222	419
	420	+ # A URL that completed on a retry missed its batch's edge purge (the
	421	+ # batch only waits so long). Late completers purge the edge themselves.
	422	+ if ($retry_count > 0) {
	423	+ Metasync_Edge_Cache_Purge::purge(array($route));
	424	+ }
	425	+
	426	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_COMPLETED, 'job', $retry_count, $outcome);
	427	+
223	428	} catch (Exception $e) {
224		- $max_retries = defined('METASYNC_OTTO_JOB_MAX_RETRIES') ? METASYNC_OTTO_JOB_MAX_RETRIES : 3;
225		-
226		- if ($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());
234		- }
	429	+ $schedule_retry('exception:' . $e->getMessage());
235	430	}
236	431	}
237	432
238	433	/**
239		- * Batch handler for cache warming and edge CDN purge.
	434	+ * Batch handler for the edge CDN purge.
240	435	*
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.
	436	+ * Runs after all per-URL jobs have had time to complete. Per-URL host caches
	437	+ * are cleared and re-warmed by the individual jobs; this batched event only
	438	+ * handles the edge CDN purge, which benefits from multi-URL API calls
	439	+ * (Cloudflare, Fastly, Akamai, Sucuri, Sevella, etc.).
245	440	*
246		- * @param array $routes List of fully-qualified URLs to warm and purge.
	441	+ * URLs still retrying are stragglers: the batch waits (up to
	442	+ * METASYNC_OTTO_BATCH_MAX_WAITS reschedules 60s apart) so their freshly
	443	+ * synced content — not stale content — reaches the edge. URLs that complete
	444	+ * after the wait budget purge the edge themselves from the retry path.
	445	+ *
	446	+ * @param array $routes List of fully-qualified URLs in this batch.
	447	+ * @param int $wait_count How many times this batch has already waited.
247	448	*/
248		-function metasync_handle_otto_batch_cache_job($routes) {
	449	+function metasync_handle_otto_batch_cache_job($routes = array(), $wait_count = 0) {
	450	+ // WP-Cron events persist independently of plugin code. A stale or malformed
	451	+ // event may therefore invoke this callback without its required route list.
	452	+ // Docblock type isn't enforced at runtime; a stale cron record can pass a non-array.
	453	+ // @phpstan-ignore-next-line function.alreadyNarrowedType
	454	+ if (!is_array($routes) || empty($routes)) {
	455	+ 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
	456	+ return;
	457	+ }
	458	+
	459	+ $max_waits = defined('METASYNC_OTTO_BATCH_MAX_WAITS') ? METASYNC_OTTO_BATCH_MAX_WAITS : 3;
	460	+ // Docblock type isn't enforced at runtime; a stale cron record can pass a non-numeric.
	461	+ // @phpstan-ignore-next-line function.alreadyNarrowedType
	462	+ $wait_count = is_numeric($wait_count) ? (int) $wait_count : 0;
	463	+
	464	+ # Split the batch: terminally-done URLs vs. URLs with work still in
	465	+ # flight. Unknown-state URLs (legacy events scheduled before the status
	466	+ # store existed, or deduped duplicates whose entry expired) count as done
	467	+ # — their per-URL job has either finished or never will, and both mean
	468	+ # the edge purge should not stall behind them.
	469	+ $ready = array();
	470	+ $stragglers = 0;
	471	+ foreach ($routes as $route) {
	472	+ $entry = Metasync_Otto_Job_Status::get($route);
	473	+ $state = is_array($entry) && isset($entry['state']) ? $entry['state'] : null;
	474	+ if ($state === Metasync_Otto_Job_Status::STATE_RETRYING
	475	+ || $state === Metasync_Otto_Job_Status::STATE_ACCEPTED) {
	476	+ $stragglers++;
	477	+ } else {
	478	+ # completed / failed / unknown: nothing left to wait for.
	479	+ $ready[] = $route;
	480	+ }
	481	+ }
	482	+
	483	+ # Still live work in the batch and wait budget left: hold the edge purge.
	484	+ if ($stragglers > 0 && $wait_count < $max_waits) {
	485	+ metasync_otto_schedule_single_event(time() + 60, 'metasync_process_otto_batch_cache_job', array($routes, $wait_count + 1));
	486	+ return;
	487	+ }
	488	+
	489	+ if (empty($ready)) {
	490	+ return;
	491	+ }
	492	+
249	493	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($routes);
255		-
256	494	# Purge edge CDN caches (Cloudflare, Fastly, Akamai, Sucuri, Sevalla, etc.)
257	495	# Tag-based providers purge only the affected posts; full-flush providers fire once per batch.
258		- Metasync_Edge_Cache_Purge::purge($routes);
	496	+ Metasync_Edge_Cache_Purge::purge($ready);
259	497	} catch (Exception $e) {
260	498	metasync_record_failed_action('metasync_process_otto_batch_cache_job');
261		- error_log('MetaSync OTTO: batch cache job failed for ' . count($routes) . ' URLs: ' . $e->getMessage() . ' in ' . $e->getFile() . ':' . $e->getLine());
	499	+ error_log('MetaSync OTTO: batch cache job failed for ' . count($ready) . ' URLs: ' . $e->getMessage() . ' in ' . $e->getFile() . ':' . $e->getLine()); // phpcs:ignore WordPress.PHP.DevelopmentFunctions.error_log_error_log -- genuine failure path, bounded, no secrets
262	500	}
263	501	}
264	502
265		-# NOTE: Cache system removed - these functions are no longer needed
266		-# Kept for backward compatibility in case old cache directories need cleanup
267		-function metasync_deleteDir($dir) {
268		- if (!is_dir($dir)) {
269		- return false;
	503	+/**
	504	+ * Drainer: work off URLs that could not be scheduled inline (batch overflow
	505	+ * or a cron that refused the event) from the durable pending queue.
	506	+ *
	507	+ * Processes at most 25 URLs per run to stay friendly to WP-Cron, schedules
	508	+ * the same batched edge purge as the webhook path for the chunk it handled,
	509	+ * and reschedules itself while the queue is non-empty. URLs it takes off
	510	+ * the queue are scheduled unconditionally unless a real cron event already
	511	+ * exists for them — the webhook's overflow bookkeeping must never veto,
	512	+ * because by then the URL has already left the durable queue.
	513	+ */
	514	+function metasync_handle_otto_pending_drainer() {
	515	+ $chunk = Metasync_Otto_Job_Status::take_pending(25);
	516	+ if (empty($chunk)) {
	517	+ return;
270	518	}
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);
	519	+
	520	+ $now = time();
	521	+ $scheduled = 0;
	522	+ foreach ($chunk as $key => $route) {
	523	+ # This run owns the chunk: take_pending() already removed these URLs
	524	+ # from the durable queue, so skipping here would drop them. Only a
	525	+ # real cron event vetoes — the webhook's overflow ledger row must
	526	+ # not (see metasync_otto_job_event_pending).
	527	+ if (metasync_otto_job_event_pending($route)) {
	528	+ continue;
	529	+ }
	530	+ if (metasync_otto_schedule_single_event($now + $key, 'metasync_process_otto_crawl_url_job', array($route))) {
	531	+ Metasync_Otto_Job_Status::record($route, Metasync_Otto_Job_Status::STATE_ACCEPTED, 'drain', 0, '');
	532	+ $scheduled++;
276	533	} else {
277		- unlink($filePath);
	534	+ # Cron still refusing: put it back at the end of the queue.
	535	+ Metasync_Otto_Job_Status::enqueue_pending($route);
278	536	}
279	537	}
280		- return rmdir($dir);
281		-}
282	538
283		-# Cleanup function for removing old cache directories (if they exist)
284		-function metasync_invalidate_all_caches($folder = ''){
285		- # Cache system removed - this function only exists to clean up old cache directories
286		- if(!defined('WP_CONTENT_DIR')){
287		- return false;
	539	+ if ($scheduled > 0) {
	540	+ metasync_otto_schedule_single_event($now + count($chunk) + 5, 'metasync_process_otto_batch_cache_job', array($chunk));
288	541	}
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 . '/' . $folder;
	542	+
	543	+ if (Metasync_Otto_Job_Status::pending_count() > 0) {
	544	+ metasync_otto_schedule_single_event($now + 60, 'metasync_otto_pending_drainer', array());
293	545	}
294		- if(is_dir($cache_dir)){
295		- metasync_deleteDir($cache_dir);
296		- }
297	546	}
298	547
299	548	// metasync_is_custom_or_lps_page() now lives in includes/metasync-helpers.php
300	549	// (loaded unconditionally before this file) so all SEO surfaces share one rule.
		@@ -319,15 +568,15 @@
319	568	*/
320	569	function metasync_is_elementor_editor_request() {
321	570	# The preview iframe loads the front end with ?elementor-preview=POST_ID.
322	571	# This is the request OTTO would otherwise buffer and corrupt.
323		- if (isset($_GET['elementor-preview'])) {
	572	+ if (isset($_GET['elementor-preview'])) { // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only editor request detection
324	573	return true;
325	574	}
326	575
327	576	# Elementor editor / app entry points carried as an action on the front end.
328		- if (isset($_REQUEST['action'])
329		- && in_array($_REQUEST['action'], array('elementor', 'elementor_ajax'), true)
	577	+ if (isset($_REQUEST['action']) // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only editor request detection
	578	+ && in_array($_REQUEST['action'], array('elementor', 'elementor_ajax'), true) // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only editor request detection
330	579	) {
331	580	return true;
332	581	}
333	582
		@@ -371,8 +620,24 @@
371	620	if (($_SERVER['REQUEST_METHOD'] ?? 'GET') !== 'GET') {
372	621	return;
373	622	}
374	623
	624	+ # Never run OTTO on WordPress core draft/pending previews
	625	+ # (?p=<ID>&preview=true, with preview_id/preview_nonce on the full link).
	626	+ # Previews are per-user, non-cacheable, and routed purely by query string —
	627	+ # and OTTO's route identity ignores the query string, so a processed preview
	628	+ # is answered with the route's public page (the homepage for /?p=…) instead
	629	+ # of the draft the editor asked for. Skip early and let WP render it.
	630	+ if (
	631	+ (isset($_GET['preview']) && $_GET['preview'] === 'true') || // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
	632	+ !empty($_GET['preview_id']) // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
	633	+ ) {
	634	+ if (!defined('DONOTCACHEPAGE')) {
	635	+ define('DONOTCACHEPAGE', true);
	636	+ }
	637	+ return;
	638	+ }
	639	+
375	640	# Never run OTTO on the Elementor editor/preview iframe. It is a logged-in
376	641	# front-end GET (not is_admin()), so it slips past the admin guard below;
377	642	# buffering and rewriting it breaks the editor canvas ("fails to load").
378	643	if (metasync_is_elementor_editor_request()) {
		@@ -405,9 +670,9 @@
405	670	# then lower-case and strip surrounding slashes.
406	671	$otto_req_path = explode('?', (string) wp_unslash($_SERVER['REQUEST_URI'] ?? ''), 2)[0];
407	672	$otto_req_path = strtolower(trim($otto_req_path, '/'));
408	673	# Strip the site's base path so first-segment matching works on subdir installs.
409		- $otto_home_path = trim((string) parse_url(home_url(), PHP_URL_PATH), '/');
	674	+ $otto_home_path = trim((string) wp_parse_url(home_url(), PHP_URL_PATH), '/');
410	675	if ($otto_home_path !== '' && strpos($otto_req_path, $otto_home_path . '/') === 0) {
411	676	$otto_req_path = substr($otto_req_path, strlen($otto_home_path) + 1);
412	677	}
413	678	$otto_first_segment = explode('/', $otto_req_path, 2)[0];
		@@ -425,11 +690,11 @@
425	690	# ──────────────────────────────────────────────────────────────────────
426	691
427	692	if (
428	693	# disable ajax calls
429		- isset($_GET['ucfrontajaxaction']) ||
	694	+ isset($_GET['ucfrontajaxaction']) || // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
430	695	# OTTO Preview mode - skip OTTO when previewing original content
431		- (isset($_GET['otto_preview']) && $_GET['otto_preview'] === '1') ||
	696	+ (isset($_GET['otto_preview']) && $_GET['otto_preview'] === '1') || // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
432	697	# WooCommerce shop archive page only (products and categories now use SSR)
433	698	//(function_exists('is_shop') && is_shop()) ||
434	699	# Cart page
435	700	(function_exists('is_cart') && is_cart()) ||
		@@ -441,32 +706,32 @@
441	706	(function_exists('wp_doing_ajax') && wp_doing_ajax()) ||
442	707	# check by constant
443	708	(defined('DOING_AJAX') && DOING_AJAX) ||
444	709	# WooCommerce AJAX endpoint (e.g., ?wc-ajax=update_cart)
445		- (isset($_REQUEST['wc-ajax']) && !empty($_REQUEST['wc-ajax'])) ||
	710	+ (isset($_REQUEST['wc-ajax']) && !empty($_REQUEST['wc-ajax'])) || // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
446	711	# AJAX requests via X-Requested-With header
447	712	(isset($_SERVER['HTTP_X_REQUESTED_WITH']) && strtolower($_SERVER['HTTP_X_REQUESTED_WITH']) === 'xmlhttprequest') ||
448	713	# Gravity Forms submission detection - skip OTTO to allow form processing
449		- (isset($_POST['gform_submit']) && (
450		- is_array($_POST['gform_submit']) ||
451		- (is_string($_POST['gform_submit']) && isset($_POST['is_submit_' . $_POST['gform_submit']]) && !empty($_POST['gform_submit']))
	714	+ (isset($_POST['gform_submit']) && ( // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
	715	+ is_array($_POST['gform_submit']) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
	716	+ (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
452	717	)) ||
453	718	# Gravity Forms AJAX submission
454		- (isset($_POST['gform_ajax']) && isset($_POST['gform_submit']) && (
455		- is_array($_POST['gform_submit']) || !empty($_POST['gform_submit'])
	719	+ (isset($_POST['gform_ajax']) && isset($_POST['gform_submit']) && ( // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
	720	+ is_array($_POST['gform_submit']) || !empty($_POST['gform_submit']) // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
456	721	)) ||
457	722	# Gravity Forms file upload
458		- (isset($_POST['gform_uploaded_files'])) ||
	723	+ (isset($_POST['gform_uploaded_files'])) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
459	724	# Any Gravity Forms POST parameter
460		- (isset($_POST['gform_submit']) || isset($_POST['gform_unique_id']) || isset($_POST['gform_field_values'])) ||
	725	+ (isset($_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
461	726	# Formidable Forms AJAX submission detection - skip OTTO to allow form processing
462		- (isset($_POST['action']) && $_POST['action'] === 'frm_entries_create') ||
	727	+ (isset($_POST['action']) && $_POST['action'] === 'frm_entries_create') || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
463	728	# Formidable Forms POST parameters
464		- (isset($_POST['form_id']) && !empty($_POST['form_id'])) ||
	729	+ (isset($_POST['form_id']) && !empty($_POST['form_id'])) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
465	730	# Formidable Forms action parameter
466		- (isset($_POST['frm_action']) && !empty($_POST['frm_action'])) ||
	731	+ (isset($_POST['frm_action']) && !empty($_POST['frm_action'])) || // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
467	732	# Formidable Forms item_key (used in form submissions)
468		- (isset($_POST['item_key']) && !empty($_POST['item_key']))
	733	+ (isset($_POST['item_key']) && !empty($_POST['item_key'])) // phpcs:ignore WordPress.Security.NonceVerification.Missing -- read-only front-end check that skips OTTO
469	734	) {
470	735	return;
471	736	}
472	737
		@@ -506,8 +771,28 @@
506	771	if ($request_path && preg_match('#\.xml$#i', $request_path)) {
507	772	return;
508	773	}
509	774
	775	+ # Machine-readable core endpoints. Using WordPress's own conditionals rather
	776	+ # than matching the path: they are exact, and a path pattern here would also
	777	+ # match ordinary content whose slug merely starts with the same characters
	778	+ # (e.g. /robots.txt-explained/, /favicon.ico-vs-svg/).
	779	+ # is_favicon() is WP 5.4+; this plugin supports 5.2.
	780	+ if (is_robots() || (function_exists('is_favicon') && is_favicon())) {
	781	+ return;
	782	+ }
	783	+
	784	+ # Protocol paths and static assets that are never OTTO-able content.
	785	+ # The guard fails open — if the strategy is unavailable we let the request
	786	+ # proceed rather than fatal. metasync_otto_class_provides() rather than a
	787	+ # bare class_exists(): a partially updated install can leave an older copy of
	788	+ # the strategy class loaded that predates is_non_content_path(), which
	789	+ # satisfies class_exists() and then fatals on every front-end request.
	790	+ if (metasync_otto_class_provides('Metasync_Otto_Render_Strategy', 'is_non_content_path')
	791	+ && Metasync_Otto_Render_Strategy::is_non_content_path($request_path)) {
	792	+ return;
	793	+ }
	794	+
510	795	# BOT DETECTION: Always detect bots so crawl data reaches the SA backend.
511	796	# Real bots don't execute JS, so otto-tracker.js never fires for them.
512	797	# push_crawl_log_to_sa() fires a non-blocking wp_remote_post here instead.
513	798	$bot_detector = Metasync_Otto_Bot_Detector::get_instance();
		@@ -515,15 +800,46 @@
515	800	if ( $detection['is_bot'] ) {
516	801	$bot_detector->push_crawl_log_to_sa( $detection, $current_url );
517	802	}
518	803
519		- # Throttle OTTO rendering for non-search-engine bots: at most one render
520		- # per URL+bot every 5 minutes. Humans and verified search engines
521		- # (Googlebot, Bingbot, etc.) are never throttled — their hits always get
522		- # full OTTO content so indexing remains current. SEO tools, AI scrapers,
523		- # uptime monitors, and unverified crawlers are de-duplicated so repeat
524		- # hits don't re-trigger the expensive DOM-rewriting path.
525		- if ( $detection['is_bot'] && $detection['bot_type'] !== 'search_engine' ) {
	804	+ # A headless frontend renders the public HTML, so stop after the cheap cache
	805	+ # protections and crawl telemetry, before any OTTO rendering work begins.
	806	+ #
	807	+ # Routed through metasync_headless_is_active() rather than calling the class
	808	+ # directly: Metasync_Headless_Config resolves only through the committed
	809	+ # Composer classmap, so a partially updated install can leave this newer
	810	+ # caller beside a missing or older copy with no fallback lookup path. This
	811	+ # function is hooked to `wp`, so an unguarded static call is a white screen
	812	+ # on every front-end page view. The helper treats an unresolvable class as
	813	+ # "not headless", which keeps OTTO behaving as it did before headless mode.
	814	+ # @phpstan-ignore-next-line function.alreadyNarrowedType
	815	+ $metasync_is_headless = function_exists('metasync_headless_is_active')
	816	+ ? metasync_headless_is_active()
	817	+ : (class_exists('Metasync_Headless_Config') && Metasync_Headless_Config::is_active());
	818	+
	819	+ if ($metasync_is_headless) {
	820	+ return;
	821	+ }
	822	+
	823	+ # Throttle OTTO rendering for bots that are not exempt: at most one render
	824	+ # per URL+bot every 5 minutes. Humans, search engines (Googlebot, Bingbot,
	825	+ # Yahoo's Slurp, Sogou, Exabot, Applebot, ...), host cache warmers and
	826	+ # synthetic performance auditors are never throttled — their hits always get
	827	+ # full OTTO content, so indexing stays current, the page cache can actually
	828	+ # be filled, and audit tools measure what a real visitor sees. SEO tools,
	829	+ # social crawlers, archivers, AI scrapers, uptime monitors and unrecognized
	830	+ # crawlers are de-duplicated so repeat hits don't re-trigger the expensive
	831	+ # DOM-rewriting path.
	832	+ #
	833	+ # Two complementary gates, deliberately kept separate:
	834	+ # - is_unthrottled_category() consults the classifier's explicit per-bot
	835	+ # category map, so the exempt set is defined in exactly one place.
	836	+ # - metasync_otto_is_unthrottled_infrastructure_agent() matches the raw
	837	+ # user-agent (plus a query marker and a site filter), which catches
	838	+ # preload agents that have no entry in the map at all.
	839	+ if ( $detection['is_bot']
	840	+ && ! Metasync_Otto_Bot_Detector::is_unthrottled_category( $detection['bot_type'] )
	841	+ && ! metasync_otto_is_unthrottled_infrastructure_agent( $detection ) ) {
526	842	$normalized_url = strtok( $current_url, '?' );
527	843	if ( $normalized_url === false ) {
528	844	$normalized_url = $current_url;
529	845	}
		@@ -542,8 +858,27 @@
542	858	}
543	859	set_transient( $render_throttle_key, 1, 5 * MINUTE_IN_SECONDS );
544	860	}
545	861
	862	+ # Skip OTTO on 404s.
	863	+ #
	864	+ # OTTO suggestions are keyed to real, crawled pages. Asking the API about a URL
	865	+ # that returns 404 spends a full API_TIMEOUT of PHP-FPM worker time on a
	866	+ # guaranteed-empty answer. Not theoretical: in one production sample, 21 of 54
	867	+ # crawler requests were 404s on URLs that never existed on the site, each
	868	+ # costing ~2s of a worker.
	869	+ #
	870	+ # Deliberately placed AFTER bot detection so push_crawl_log_to_sa() still
	871	+ # reports the crawl to SearchAtlas — that telemetry is how the crawler side
	872	+ # can learn to stop requesting these URLs. Only the expensive suggestions
	873	+ # lookup is skipped.
	874	+ #
	875	+ # Filterable because "should a 404 page get OTTO treatment" is arguably a site
	876	+ # decision, even though there is no known case where it should.
	877	+ if (is_404() && apply_filters('metasync_otto_skip_on_404', true)) {
	878	+ return;
	879	+ }
	880	+
546	881	# Optionally skip OTTO processing for bot traffic (when the setting is enabled)
547	882	if ($bot_detector->should_skip_otto()) {
548	883	// Log the bot locally and count the saved API call
549	884	$bot_detector->log_detection($detection);
		@@ -567,9 +902,9 @@
567	902	# ?et_blog = Divi AJAX pagination callback
568	903	# /page/N/ = paginated blog/archive pages — OTTO's buffer/HTTP render causes
569	904	# module numbering mismatch between page 1 (with TB template) and page N
570	905	# (without TB template), breaking Divi's JS pagination selector matching.
571		- if (isset($_GET['et_blog']) || (is_paged() && !is_singular())) {
	906	+ if (isset($_GET['et_blog']) || (is_paged() && !is_singular())) { // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
572	907	return;
573	908	}
574	909
575	910	# Skip OTTO on search-results pages. A search request carries its
		@@ -584,9 +919,9 @@
584	919	# correct pre-rendered search page to everyone else. Search-results pages
585	920	# have no per-canonical-URL OTTO suggestions of their own, so OTTO must never
586	921	# run here. The raw $_GET['s'] check is a defensive fallback for setups where
587	922	# the main query is altered before this point.
588		- if (is_search() || !empty($_GET['s'])) {
	923	+ if (is_search() || !empty($_GET['s'])) { // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only front-end check that skips OTTO
589	924	return;
590	925	}
591	926
592	927	# Handle cache plugin compatibility early - before any caching happens
		@@ -592,9 +927,14 @@
592	927	# Handle cache plugin compatibility early - before any caching happens
593	928	metasync_otto_handle_cache_compatibility();
594	929
595	930	# check if OTTO is disabled for this specific page/post
596		- $post_id = get_the_ID();
	931	+ # Gate on is_singular() for the same reason the custom/LPS check below does:
	932	+ # on an archive, search or term query get_the_ID() returns the first post of
	933	+ # the loop — WP::register_globals() seeds $GLOBALS['post'] with it — not the
	934	+ # page being rendered. Without the gate, one excluded or per-page-disabled
	935	+ # post silently takes OTTO off every archive that happens to list it first.
	936	+ $post_id = is_singular() ? ( get_queried_object_id() ?: get_the_ID() ) : 0;
597	937	if ($post_id && class_exists('Metasync_Otto_Frontend_Toolbar')) {
598	938	if (Metasync_Otto_Frontend_Toolbar::is_otto_disabled($post_id)) {
599	939	return;
600	940	}
		@@ -611,9 +951,9 @@
611	951	# custom/LPS page; gate on is_singular() so an archive/search/term query
612	952	# can never have its object id mistaken for a custom page's post id.
613	953	$custom_page_id = is_singular() ? ( get_queried_object_id() ?: get_the_ID() ) : 0;
614	954	if (metasync_is_custom_or_lps_page($custom_page_id)) {
615		- if (!headers_sent()) {
	955	+ if (!headers_sent() && Metasync_Otto_Render_Strategy::diagnostics_enabled()) {
616	956	header('X-MetaSync-OTTO-Method: EXCLUDED');
617	957	}
618	958	return;
619	959	}
		@@ -618,15 +958,32 @@
618	958	return;
619	959	}
620	960
621	961	# check if we are having an otto request
622		- if(!empty($_GET['is_otto_page_fetch'])){
	962	+ #
	963	+ # Recognise our own internal fetch by header/User-Agent as well as by the
	964	+ # query parameter. A redirect that rebuilds the URL strips the parameter,
	965	+ # and without this the redirected request re-enters OTTO and fires another
	966	+ # internal fetch - an unbounded self-request loop.
	967	+ # method_exists() is redundant to static analysis - this MR adds the method,
	968	+ # so PHPStan proves the call always true. It is kept for the upgrade window:
	969	+ # during a plugin update an opcache can still hold the previous
	970	+ # Metasync_Otto_Render_Strategy, where class_exists() passes but the method
	971	+ # is absent, and an unguarded static call would fatal the page.
	972	+ # @phpstan-ignore-next-line function.alreadyNarrowedType
	973	+ $otto_has_fetch_detector = class_exists('Metasync_Otto_Render_Strategy') && method_exists('Metasync_Otto_Render_Strategy', 'is_internal_fetch');
623	974
	975	+ $otto_is_internal_fetch = $otto_has_fetch_detector
	976	+ ? Metasync_Otto_Render_Strategy::is_internal_fetch()
	977	+ : !empty($_GET['is_otto_page_fetch']); // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only loopback marker, request-scoped filters only
	978	+
	979	+ if($otto_is_internal_fetch){
	980	+
624	981	# Block SEO plugins NOW for this internal fetch request
625	982	# metasync_otto_block_seo_plugins();
626	983	# $_SERVER['REQUEST_URI'] = remove_query_arg('is_otto_page_fetch', $_SERVER['REQUEST_URI']);
627		- $block_title = !empty($_GET['otto_block_title']) && $_GET['otto_block_title'] === '1';
628		- $block_description = !empty($_GET['otto_block_desc']) && $_GET['otto_block_desc'] === '1';
	984	+ $block_title = !empty($_GET['otto_block_title']) && $_GET['otto_block_title'] === '1'; // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only loopback marker, request-scoped filters only
	985	+ $block_description = !empty($_GET['otto_block_desc']) && $_GET['otto_block_desc'] === '1'; // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only loopback marker, request-scoped filters only
629	986
630	987	# Block SEO plugins conditionally based on what Otto has
631	988	if ($block_title || $block_description) {
632	989	metasync_otto_block_seo_plugins($block_title, $block_description);
		@@ -662,8 +1019,19 @@
662	1019
663	1020	# get the otto uuid
664	1021	$otto_uuid = Metasync_Otto_Config::get_otto_uuid();
665	1022
	1023	+ # Upgrade-window guard: the bundled DOM parser resolves only through the
	1024	+ # otto/vendor autoloader required at the top of this file. While an update
	1025	+ # is mid-copy (or a stale opcache classmap is still resident), that
	1026	+ # dependency can briefly be undefined here, and constructing the render
	1027	+ # stack below would throw a class-not-found Error and take the whole page
	1028	+ # down with it. Degrade to an OTTO-less page for this request instead —
	1029	+ # the next request, on the settled tree, renders normally.
	1030	+ if ( ! class_exists( 'simplehtmldom\HtmlDocument' ) ) {
	1031	+ return;
	1032	+ }
	1033	+
666	1034	# start the class
667	1035	$otto = new Metasync_otto_pixel($otto_uuid);
668	1036
669	1037	# call render
		@@ -695,9 +1063,9 @@
695	1063	if ($cached !== false) {
696	1064	$has_brizy_posts = ($cached === 'yes');
697	1065	} else {
698	1066	# Query database only if cache missed
699		- $has_brizy_posts = $wpdb->get_var(
	1067	+ $has_brizy_posts = $wpdb->get_var( // phpcs:ignore WordPress.DB.DirectDatabaseQuery -- feature-detection probe for Brizy posts — postmeta flag scan has no WordPress API
700	1068	"SELECT COUNT(*) FROM {$wpdb->postmeta}
701	1069	WHERE meta_key = 'brizy_post_uid'
702	1070	AND meta_value != ''
703	1071	LIMIT 1"
		@@ -780,8 +1148,314 @@
780	1148	}
781	1149	}
782	1150
783	1151	/**
	1152	+ * How long a response that OTTO could not optimize may be cached, in seconds.
	1153	+ *
	1154	+ * Deliberately short rather than zero — see
	1155	+ * metasync_otto_limit_response_cache() for why.
	1156	+ */
	1157	+if (!defined('METASYNC_OTTO_FALLBACK_CACHE_TTL')) {
	1158	+ define('METASYNC_OTTO_FALLBACK_CACHE_TTL', 60);
	1159	+}
	1160	+
	1161	+/**
	1162	+ * Cap how long the current response may be cached, because OTTO could not apply
	1163	+ * its suggestions to it.
	1164	+ *
	1165	+ * OTTO fails open: when suggestions are unavailable, or when the HTML rewrite
	1166	+ * fails, the site's own un-optimized markup is served instead. That markup is
	1167	+ * valid — the harm is that a caching layer then stores it and serves it to
	1168	+ * everyone for the rest of its TTL (a host edge cache with s-maxage=86400 pins
	1169	+ * it for a day), which surfaces to customers as the OTTO title "reverting".
	1170	+ *
	1171	+ * A short TTL is used rather than `no-store`, which is the obvious choice but the
	1172	+ * wrong one. Some failures look transient and are not: a page that always breaks
	1173	+ * the rewriter, or a host that cannot fetch its own pages, fails identically on
	1174	+ * every request. `no-store` would make those permanently uncacheable, sending
	1175	+ * 100% of their traffic to the origin forever in exchange for markup that will
	1176	+ * never improve. A one-minute ceiling fixes the reported bug just as completely —
	1177	+ * a day-long pin becomes a minute — while capping the cost of being wrong about
	1178	+ * which failures recover.
	1179	+ *
	1180	+ * Different layers read different signals, so all of them are emitted:
	1181	+ *
	1182	+ * - Cache-Control CDNs, host edge caches, browsers. `max-age=0` keeps
	1183	+ * browsers revalidating while `s-maxage` lets shared
	1184	+ * caches absorb a burst for the ceiling's duration.
	1185	+ * - X-Accel-Expires Nginx's own directive. Evaluated with higher
	1186	+ * precedence than Cache-Control, and not covered by
	1187	+ * the `fastcgi_ignore_headers Cache-Control Expires`
	1188	+ * recipe common on managed hosts — so it can survive
	1189	+ * a config that discards Cache-Control. Nginx strips
	1190	+ * X-Accel-* before the response reaches the client,
	1191	+ * so there is no visitor-facing side effect.
	1192	+ * - Surrogate-Control Varnish and Fastly. Note that on stock Varnish the
	1193	+ * mere presence of this header suppresses its
	1194	+ * Cache-Control check, so it must carry a real
	1195	+ * lifetime rather than be left at some other value.
	1196	+ * - DONOTCACHEPAGE WordPress-level page caches (WP Rocket, W3TC,
	1197	+ * LiteSpeed, SG Optimizer). These run inside PHP and
	1198	+ * never see response headers. They have no notion of a
	1199	+ * short TTL — it is cache or do not — so they skip
	1200	+ * entirely. That is safe here because callers only
	1201	+ * invoke this for failures expected to resolve; a
	1202	+ * permanent one (an oversized page, a host that cannot
	1203	+ * reach itself) is left fully cacheable by the caller
	1204	+ * and never reaches this function.
	1205	+ *
	1206	+ * Cache-tagging headers (Surrogate-Key, Cache-Tag, Edge-Cache-Tag) are
	1207	+ * deliberately left in place. They are purge-targeting labels, not cacheability
	1208	+ * directives — no CDN stores a response *because* a tag is present. On a layer
	1209	+ * that stores this response anyway, those tags are the only handle a later purge
	1210	+ * has on it, so removing them would discard the recovery path exactly where it
	1211	+ * is needed most.
	1212	+ *
	1213	+ * Some layers may honour none of this; a managed host that overrides origin
	1214	+ * cache headers will cache the response regardless. This bounds the damage where
	1215	+ * it can and is inert where it cannot.
	1216	+ *
	1217	+ * @param string $reason Short machine-readable reason, surfaced as a response
	1218	+ * header for debugging (e.g. 'API_ERROR', 'RENDER_FAILED').
	1219	+ */
	1220	+/**
	1221	+ * Build the response headers that cap how long an un-optimized response may be
	1222	+ * cached.
	1223	+ *
	1224	+ * Thin wrapper over Metasync_Otto_Render_Strategy::fallback_cache_headers(), which
	1225	+ * owns the header set so it can be unit-tested (this file cannot be loaded in a
	1226	+ * test context). Returns an empty array if the strategy class is somehow absent,
	1227	+ * so a caller emits nothing rather than an untested divergent copy.
	1228	+ *
	1229	+ * @param int $ttl Cache ceiling in seconds.
	1230	+ * @param string $reason Optional machine-readable reason for the debug header.
	1231	+ * @return array Header name => value.
	1232	+ */
	1233	+function metasync_otto_fallback_cache_headers($ttl, $reason = '') {
	1234	+ if (!class_exists('Metasync_Otto_Render_Strategy')) {
	1235	+ return [];
	1236	+ }
	1237	+
	1238	+ return Metasync_Otto_Render_Strategy::fallback_cache_headers_for_response($ttl, $reason);
	1239	+}
	1240	+
	1241	+function metasync_otto_limit_response_cache($reason = '') {
	1242	+ # This function must only ever make the response LESS cacheable. header()
	1243	+ # replaces by default and the capped set carries X-Accel-Expires, which nginx
	1244	+ # evaluates ahead of Cache-Control — so emitting it unconditionally could
	1245	+ # override an upstream no-store and grant edge caching that nothing had
	1246	+ # granted. The strategy reconciles existing cache directives before emitting
	1247	+ # a capped set.
	1248	+
	1249	+ $strategy_available = class_exists('Metasync_Otto_Render_Strategy');
	1250	+
	1251	+ # 1. Logged-in users. OTTO renders for them unless the "disable on logged in"
	1252	+ # setting is switched on, which is off by default — so this is the common
	1253	+ # case, not an edge case. Their page may carry session-specific or
	1254	+ # membership-gated content, and capping at 60 seconds would still let a
	1255	+ # shared cache store and re-serve it to somebody else.
	1256	+ if (function_exists('is_user_logged_in') && is_user_logged_in()) {
	1257	+ if (!headers_sent() && $strategy_available) {
	1258	+ foreach (Metasync_Otto_Render_Strategy::no_store_headers($reason) as $name => $value) {
	1259	+ header($name . ': ' . $value);
	1260	+ }
	1261	+ }
	1262	+ if (!defined('DONOTCACHEPAGE')) {
	1263	+ define('DONOTCACHEPAGE', true);
	1264	+ }
	1265	+ return;
	1266	+ }
	1267	+
	1268	+ # 2. Something upstream already restricted this response. Leave its headers
	1269	+ # alone — a page deliberately marked no-store must not be handed a 60
	1270	+ # second shared-cache lifetime by us.
	1271	+ # The strategy turns an existing no-store into a consistent no-store set and
	1272	+ # uses the strictest known existing lifetime for every cache layer.
	1273	+ $ttl = (int) METASYNC_OTTO_FALLBACK_CACHE_TTL;
	1274	+
	1275	+ if (!headers_sent()) {
	1276	+ foreach (metasync_otto_fallback_cache_headers($ttl, $reason) as $name => $value) {
	1277	+ header($name . ': ' . $value);
	1278	+ }
	1279	+ }
	1280	+
	1281	+ # Reaches page-cache plugins, which do not see the headers above.
	1282	+ if (!defined('DONOTCACHEPAGE')) {
	1283	+ define('DONOTCACHEPAGE', true);
	1284	+ }
	1285	+}
	1286	+
	1287	+/**
	1288	+ * Report an OTTO render failure to the local error log and to Sentry.
	1289	+ *
	1290	+ * These failures were invisible: the render path swallowed them and returned the
	1291	+ * original HTML, so nobody knew how often they fired. Reporting them is what makes
	1292	+ * a persistently-failing URL discoverable instead of silently costing origin work.
	1293	+ *
	1294	+ * Two constraints shape this:
	1295	+ *
	1296	+ * 1. It runs on the visitor render path, so it must not add latency. The Sentry
	1297	+ * report is therefore queued here and sent by
	1298	+ * metasync_otto_flush_render_failure_reports() after the response has been
	1299	+ * handed to the visitor, so the cost is invisible rather than merely small.
	1300	+ *
	1301	+ * Two alternatives were tried and rejected. Sending inline with a blocking
	1302	+ * transport — cURL with a 5s connect and 5s read timeout — could add ten
	1303	+ * seconds to a page load, precisely when the request is already degraded.
	1304	+ * Sending inline non-blocking is far cheaper but still not free: opening the
	1305	+ * connection and writing the body costs the visitor something.
	1306	+ *
	1307	+ * A WP-Cron single event was the other obvious option and is worse than both:
	1308	+ * every scheduled event is stored in the `cron` option, which is autoloaded on
	1309	+ * every request site-wide, and WP's own 10-minute duplicate suppression keys on
	1310	+ * the serialized args — which carry per-request byte counts here, so it would
	1311	+ * never engage. A site with many failing URLs would grow that option
	1312	+ * unboundedly and pay for it on every page load, which is the cost the deferral
	1313	+ * was meant to avoid in the first place.
	1314	+ *
	1315	+ * The local log stays synchronous. It is cheaper than Sentry but not free —
	1316	+ * Metasync_Error_Logger::log() rewrites a summary option on each call — which is
	1317	+ * the second reason for the throttle below.
	1318	+ *
	1319	+ * 2. A failing page can be requested thousands of times an hour, so reports are
	1320	+ * throttled per reason+URL.
	1321	+ *
	1322	+ * @param string $reason Machine-readable failure reason (e.g. 'RENDER_EXCEPTION').
	1323	+ * @param string $message Human-readable description.
	1324	+ * @param array $context Extra key/value context.
	1325	+ * @param string $level Sentry level: 'error' or 'warning'.
	1326	+ */
	1327	+function metasync_otto_report_render_failure($reason, $message, $context = [], $level = 'error') {
	1328	+ $url = '';
	1329	+ if (!empty($_SERVER['HTTP_HOST']) && !empty($_SERVER['REQUEST_URI'])) {
	1330	+ $url = (is_ssl() ? 'https://' : 'http://')
	1331	+ . $_SERVER['HTTP_HOST']
	1332	+ . strtok(sanitize_text_field(wp_unslash($_SERVER['REQUEST_URI'])), '?');
	1333	+ }
	1334	+
	1335	+ # Throttle: at most one report per reason+URL per 5 minutes. Requests without a
	1336	+ # resolvable URL (CLI, odd SAPIs) collapse onto one key, which is intended —
	1337	+ # they cannot be attributed to a page anyway.
	1338	+ $throttle_key = 'metasync_otto_fail_' . md5($reason . '|' . $url);
	1339	+ if (get_transient($throttle_key)) {
	1340	+ return;
	1341	+ }
	1342	+ set_transient($throttle_key, 1, 5 * MINUTE_IN_SECONDS);
	1343	+
	1344	+ $context = array_merge([
	1345	+ 'reason' => $reason,
	1346	+ 'url' => $url,
	1347	+ ], $context);
	1348	+
	1349	+ # Local structured log. Stays on the customer's site, so it is useful to
	1350	+ # support but not visible to us — hence the Sentry dispatch below.
	1351	+ if (class_exists('Metasync_Error_Logger')) {
	1352	+ Metasync_Error_Logger::log(
	1353	+ Metasync_Error_Logger::CATEGORY_OTTO_RENDER,
	1354	+ $level === 'warning'
	1355	+ ? Metasync_Error_Logger::SEVERITY_WARNING
	1356	+ : Metasync_Error_Logger::SEVERITY_ERROR,
	1357	+ $message,
	1358	+ $context
	1359	+ );
	1360	+ }
	1361	+
	1362	+ # Sentry. Queued rather than sent here: even a non-blocking dispatch costs the
	1363	+ # visitor something (opening the connection and writing the body), and this runs
	1364	+ # on a request that is already degraded. The queue is flushed after the response
	1365	+ # has been handed to the visitor — see metasync_otto_flush_render_failure_reports().
	1366	+ if (!function_exists('metasync_sentry_capture_message_nonblocking')) {
	1367	+ return;
	1368	+ }
	1369	+
	1370	+ $GLOBALS['metasync_otto_pending_failure_reports'][] = [
	1371	+ 'message' => $message,
	1372	+ 'level' => $level,
	1373	+ 'context' => $context,
	1374	+ ];
	1375	+
	1376	+ if (empty($GLOBALS['metasync_otto_failure_flush_hooked'])) {
	1377	+ $GLOBALS['metasync_otto_failure_flush_hooked'] = true;
	1378	+ # PHP_INT_MAX so this runs after everything else registered on shutdown —
	1379	+ # WordPress flushes its own output buffers and closes the object cache at the
	1380	+ # default priority, and this handler ends output for the request.
	1381	+ add_action('shutdown', 'metasync_otto_flush_render_failure_reports', PHP_INT_MAX);
	1382	+ }
	1383	+}
	1384	+
	1385	+/**
	1386	+ * Send any queued render-failure reports after the response has been delivered.
	1387	+ *
	1388	+ * fastcgi_finish_request() (PHP-FPM) and litespeed_finish_request() (LiteSpeed)
	1389	+ * flush the response and close the connection to the visitor while leaving PHP
	1390	+ * running. Calling one of them here means the page is already on its way before
	1391	+ * any telemetry work starts, so the reporting cost is invisible to the visitor
	1392	+ * rather than merely small.
	1393	+ *
	1394	+ * Neither exists under mod_php or the CLI SAPI. There the reports are still sent
	1395	+ * at the very end of the request, after all page output has been generated and
	1396	+ * flushed — the same ordering, just without the connection being closed first.
	1397	+ *
	1398	+ * The dispatch itself stays non-blocking. Now that the visitor is served, the only
	1399	+ * remaining cost is PHP worker occupancy, and a blocking transport with a 5s
	1400	+ * connect and 5s read timeout could tie up a worker for ten seconds per report on
	1401	+ * a site that is failing often.
	1402	+ */
	1403	+function metasync_otto_flush_render_failure_reports() {
	1404	+ $reports = $GLOBALS['metasync_otto_pending_failure_reports'] ?? [];
	1405	+ $GLOBALS['metasync_otto_pending_failure_reports'] = [];
	1406	+
	1407	+ if (empty($reports)) {
	1408	+ return;
	1409	+ }
	1410	+
	1411	+ if (function_exists('fastcgi_finish_request')) {
	1412	+ fastcgi_finish_request();
	1413	+ } elseif (function_exists('litespeed_finish_request')) {
	1414	+ litespeed_finish_request();
	1415	+ }
	1416	+
	1417	+ if (!function_exists('metasync_sentry_capture_message_nonblocking')) {
	1418	+ return;
	1419	+ }
	1420	+
	1421	+ foreach ($reports as $report) {
	1422	+ try {
	1423	+ metasync_sentry_capture_message_nonblocking(
	1424	+ $report['message'],
	1425	+ $report['level'],
	1426	+ $report['context']
	1427	+ );
	1428	+ } catch (Exception $e) {
	1429	+ # Telemetry must never be able to break a request.
	1430	+ } catch (Error $e) {
	1431	+ # Same.
	1432	+ }
	1433	+ }
	1434	+}
	1435	+
	1436	+/**
	1437	+ * Whether MetaSync is still emitting Open Graph / Twitter tags.
	1438	+ *
	1439	+ * The Social Media & Open Graph switch stops MetaSync writing those tags on
	1440	+ * every render path, so the suppression below has to stand down with it —
	1441	+ * removing another plugin's og:/twitter: tag while contributing none of our own
	1442	+ * would leave the page with no social tags at all.
	1443	+ *
	1444	+ * The class_exists() guard keeps this usable if the flags file has not loaded
	1445	+ * yet; treating that window as "enabled" preserves the previous behaviour.
	1446	+ *
	1447	+ * @return bool
	1448	+ */
	1449	+function metasync_otto_social_output_enabled() {
	1450	+ if (!class_exists('Metasync_Feature_Flags')) {
	1451	+ return true;
	1452	+ }
	1453	+
	1454	+ return Metasync_Feature_Flags::is_enabled(Metasync_Feature_Flags::SOCIAL_OG);
	1455	+}
	1456	+
	1457	+/**
784	1458	* Block SEO plugins conditionally based on what Otto is providing
785	1459	* Only blocks title if Otto has title, only blocks description if Otto has description
786	1460	* This prevents duplicate SEO tags while allowing fallback to SEO plugins when Otto has no data.
787	1461	* Supports Yoast SEO, Rank Math, and AIOSEO (free + pro).
		@@ -821,11 +1495,19 @@
821	1495
822	1496
823	1497	# Remove description presenters only when OTTO has a description
824	1498	if ($block_description) {
	1499	+ # The plain meta description belongs to the SEO title/description
	1500	+ # feature, which has its own setting.
825	1501	$presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Meta_Description_Presenter';
826		- $presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Open_Graph\Description_Presenter';
827		- $presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Twitter\Description_Presenter';
	1502	+
	1503	+ # The OG/Twitter descriptions are social tags. With that feature
	1504	+ # switched off MetaSync writes none of its own, so Yoast's must be
	1505	+ # left alone rather than removed with nothing to replace them.
	1506	+ if (metasync_otto_social_output_enabled()) {
	1507	+ $presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Open_Graph\Description_Presenter';
	1508	+ $presenters_to_remove[] = 'Yoast\WP\SEO\Presenters\Twitter\Description_Presenter';
	1509	+ }
828	1510	}
829	1511
830	1512	foreach ($presenters as $key => $presenter) {
831	1513	// Safely get class name, suppressing autoload errors
		@@ -843,11 +1525,13 @@
843	1525	# Disable Rank Math
844	1526	if (is_plugin_active('seo-by-rank-math/rank-math.php') ||
845	1527	is_plugin_active('seo-by-rankmath/rank-math.php')) {
846	1528
847		- if ($block_title) {
848		- add_filter('rank_math/frontend/title', '__return_empty_string', 999);
849		- }
	1529	+ # TITLE: Never blank Rank Math's title output during SSR fetch.
	1530	+ # rank_math/frontend/title also gates Rank Math's og:title/twitter:title
	1531	+ # output, so forcing it empty blanks those social tags along with the
	1532	+ # document title. Let Rank Math render its own title/og/twitter tags;
	1533	+ # OTTO's replace_title() and the dedup/precedence pass clean up duplicates.
850	1534
851	1535	if ($block_description) {
852	1536	add_filter('rank_math/frontend/description', '__return_false', 999);
853	1537	add_filter('rank_math/frontend/show_keywords', '__return_false', 999);
		@@ -858,17 +1542,22 @@
858	1542	if (is_plugin_active('all-in-one-seo-pack/all_in_one_seo_pack.php') ||
859	1543	is_plugin_active('all-in-one-seo-pack-pro/all_in_one_seo_pack.php')) {
860	1544
861	1545	if ($block_title) {
	1546	+ # The document title belongs to the SEO title feature, which has its
	1547	+ # own setting; only the og:/twitter: unsets below are social.
862	1548	add_filter('aioseo_title', '__return_empty_string', 999);
863		- add_filter('aioseo_facebook_tags', function($meta) {
864		- if (is_array($meta)) { unset($meta['og:title']); }
865		- return $meta;
866		- }, 999);
867		- add_filter('aioseo_twitter_tags', function($meta) {
868		- if (is_array($meta)) { unset($meta['twitter:title']); }
869		- return $meta;
870		- }, 999);
	1549	+
	1550	+ if (metasync_otto_social_output_enabled()) {
	1551	+ add_filter('aioseo_facebook_tags', function($meta) {
	1552	+ if (is_array($meta)) { unset($meta['og:title']); }
	1553	+ return $meta;
	1554	+ }, 999);
	1555	+ add_filter('aioseo_twitter_tags', function($meta) {
	1556	+ if (is_array($meta)) { unset($meta['twitter:title']); }
	1557	+ return $meta;
	1558	+ }, 999);
	1559	+ }
871	1560	}
872	1561
873	1562	if ($block_description) {
874	1563	# Use granular tag list when available to only block what OTTO provides
		@@ -873,10 +1562,12 @@
873	1562	if ($block_description) {
874	1563	# Use granular tag list when available to only block what OTTO provides
875	1564	$tags = !empty($description_tags) ? $description_tags : [];
876	1565	$block_standard = empty($tags) || in_array('meta[name=description]', $tags);
877		- $block_og_desc = empty($tags) || in_array('meta[property=og:description]', $tags);
878		- $block_tw_desc = empty($tags) || in_array('meta[name=twitter:description]', $tags);
	1566	+ # Social tags only get blocked while MetaSync is still emitting them.
	1567	+ $social_enabled = metasync_otto_social_output_enabled();
	1568	+ $block_og_desc = $social_enabled && (empty($tags) || in_array('meta[property=og:description]', $tags));
	1569	+ $block_tw_desc = $social_enabled && (empty($tags) || in_array('meta[name=twitter:description]', $tags));
879	1570
880	1571	if ($block_standard) {
881	1572	add_filter('aioseo_description', '__return_empty_string', 999);
882	1573	}
		@@ -974,8 +1665,17 @@
974	1665
975	1666	# Handle AJAX Clear Cache request
976	1667	# NOTE: Cache system removed - this is now a no-op
977	1668	function metasync_clear_otto_cache_handler() {
	1669	+ # State-changing admin action: verify the nonce sent by admin/js
	1670	+ # (clear_otto_caches) and confirm plugin access before clearing the
	1671	+ # detection transient or queueing the loopback re-check.
	1672	+ check_ajax_referer('metasync_clear_otto_cache', 'nonce');
	1673	+
	1674	+ if (!Metasync::current_user_has_plugin_access()) {
	1675	+ wp_send_json_error(['message' => 'Not authorized']);
	1676	+ }
	1677	+
978	1678	if (!empty($_GET['clear_otto_cache'])) {
979	1679	delete_transient('metasync_otto_js_detected');
980	1680	# Re-run the JS detection in the background so the notice refreshes.
981	1681	metasync_schedule_otto_js_check();
		@@ -1019,15 +1719,21 @@
1019	1719	add_action('wp', 'metasync_start_otto');
1020	1720
1021	1721	# ENHANCED OTTO SEO INTEGRATION
1022	1722	# Register async SEO processing hook
1023		- add_action('metasync_process_seo_job', 'metasync_process_otto_seo_data', 10, 3);
	1723	+ add_action('metasync_process_seo_job', 'metasync_process_otto_seo_data', 10, 4);
1024	1724	add_action('metasync_process_otto_crawl_url_job', 'metasync_handle_otto_crawl_url_job', 10, 2);
1025		- add_action('metasync_process_otto_batch_cache_job', 'metasync_handle_otto_batch_cache_job');
	1725	+ add_action('metasync_process_otto_batch_cache_job', 'metasync_handle_otto_batch_cache_job', 10, 2);
	1726	+ add_action('metasync_otto_pending_drainer', 'metasync_handle_otto_pending_drainer');
1026	1727
1027	1728	# Process OTTO SEO data and update WordPress meta fields for SEO plugins
1028	1729	# This function now runs asynchronously via WordPress cron system
1029	1730	#
	1731	+ # Returns one of the Metasync_Otto_Job_Status::OUTCOME_* strings so callers
	1732	+ # can distinguish success, confirmed no-change, retryable failure, permanent
	1733	+ # failure, and CPU deferral — a bare boolean cannot drive correct retry and
	1734	+ # purge decisions.
	1735	+ #
1030	1736	# @param string $route Fully-qualified URL to process.
1031	1737	# @param bool $allow_defer When false, skip CPU-deferral rescheduling (used when
1032	1738	# called synchronously from crawl_url_job whose own retry
1033	1739	# mechanism already handles failures).
		@@ -1032,10 +1738,14 @@
1032	1738	# called synchronously from crawl_url_job whose own retry
1033	1739	# mechanism already handles failures).
1034	1740	# @param int $deferral_count How many times this job has already been deferred for
1035	1741	# CPU load. Prevents infinite reschedule loops.
	1742	+ # @param array|null $prefetched_seo_data Suggestions already fetched by the transient
	1743	+ # warm step. When an array it is used as-is so the job
	1744	+ # makes exactly ONE API call per URL; the endpoint is
	1745	+ # only fetched here when no prefetch was provided.
1036	1746
1037		-function metasync_process_otto_seo_data($route, $allow_defer = true, $deferral_count = 0) {
	1747	+function metasync_process_otto_seo_data($route, $allow_defer = true, $deferral_count = 0, $prefetched_seo_data = null) {
1038	1748	# Maximum number of times a job can be deferred before it is dropped.
1039	1749	$max_deferrals = defined('METASYNC_SEO_JOB_MAX_DEFERRALS') ? METASYNC_SEO_JOB_MAX_DEFERRALS : 5;
1040	1750	$lock_key = null;
1041	1751	$lock_acquired = false;
		@@ -1042,9 +1752,9 @@
1042	1752
1043	1753	try {
1044	1754	# Validate input
1045	1755	if (empty($route) || !is_string($route)) {
1046		- return false;
	1756	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1047	1757	}
1048	1758
1049	1759	# CPU load check — defer if server is under load.
1050	1760	# Only reschedule when called from the cron hook (allow_defer=true) and
		@@ -1053,14 +1763,14 @@
1053	1763	# acquire-then-immediately-release it (which defeats concurrency protection).
1054	1764	if (!Metasync_CPU_Monitor::is_load_safe()) {
1055	1765	if ($allow_defer) {
1056	1766	if ($deferral_count < $max_deferrals) {
1057		- wp_schedule_single_event(
	1767	+ metasync_otto_schedule_single_event(
1058	1768	time() + 60,
1059	1769	'metasync_process_seo_job',
1060	1770	array($route, true, $deferral_count + 1)
1061	1771	);
1062		- return false;
	1772	+ return Metasync_Otto_Job_Status::OUTCOME_DEFERRED;
1063	1773	}
1064	1774	# deferral budget exhausted — RUN the job late instead
1065	1775	# of dropping it. The write is idempotent and bounded, and a
1066	1776	# late sync is strictly better than SEO data that never
		@@ -1066,11 +1776,12 @@
1066	1776	# late sync is strictly better than SEO data that never
1067	1777	# arrives. No new cron events are created, so the
1068	1778	# anti-pile-up guarantee is preserved. Fall through.
1069	1779	} else {
1070		- # allow_defer=false (sync path): return false without creating
1071		- # cron events — OTTO's next crawl re-triggers this URL.
1072		- return false;
	1780	+ # allow_defer=false (sync path): report retryable without creating
	1781	+ # cron events — the crawl-url job's retry mechanism (or OTTO's
	1782	+ # next crawl) re-triggers this URL.
	1783	+ return Metasync_Otto_Job_Status::OUTCOME_RETRYABLE;
1073	1784	}
1074	1785	}
1075	1786
1076	1787	# Concurrency lock: prevent multiple SEO jobs from running simultaneously.
		@@ -1079,15 +1790,15 @@
1079	1790	# processing this URL — reschedule once with a short delay instead of stacking.
1080	1791	$lock_key = 'metasync_seo_lock_' . md5($route);
1081	1792	if (get_transient($lock_key) !== false) {
1082	1793	if ($allow_defer && $deferral_count < $max_deferrals) {
1083		- wp_schedule_single_event(
	1794	+ metasync_otto_schedule_single_event(
1084	1795	time() + 30,
1085	1796	'metasync_process_seo_job',
1086	1797	array($route, true, $deferral_count + 1)
1087	1798	);
1088	1799	}
1089		- return false;
	1800	+ return Metasync_Otto_Job_Status::OUTCOME_RETRYABLE;
1090	1801	}
1091	1802	set_transient($lock_key, true, 120);
1092	1803	$lock_acquired = true;
1093	1804
		@@ -1096,16 +1807,15 @@
1096	1807
1097	1808	# Skip excluded URLs - don't process SEO data for them
1098	1809	if (metasync_is_otto_url_excluded($route)) {
1099	1810	//error_log('MetaSync OTTO: Skipping SEO processing for excluded URL: ' . $route);
1100		- return false;
	1811	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1101	1812	}
1102	1813
1103	1814	# Pre-flight 404 check: exclude URLs that would return 404 before making API call
1104	1815	if (!metasync_otto_is_url_available($route)) {
1105		- error_log("MetaSync OTTO: Skipping SEO processing for URL that would return 404: {$route}");
1106	1816	metasync_otto_auto_exclude_404_url($route);
1107		- return false;
	1817	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1108	1818	}
1109	1819
1110	1820	# Get OTTO UUID from settings
1111	1821	# OPTIMIZED: Use cached options
		@@ -1111,24 +1821,40 @@
1111	1821	# OPTIMIZED: Use cached options
1112	1822	$otto_uuid = Metasync_Otto_Config::get_otto_uuid();
1113	1823
1114	1824	if (empty($otto_uuid)) {
1115		- return false;
	1825	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1116	1826	}
1117	1827
1118	1828	# Meta descriptions are always enabled by default - no check needed
1119	1829
1120		- # Fetch SEO data from OTTO API
1121		- $seo_data = metasync_fetch_otto_seo_data($route, $otto_uuid);
1122		-
1123		- if (!$seo_data) {
1124		- metasync_record_failed_action( 'metasync_process_seo_job' );
1125		- return false;
	1830	+ # Obtain the OTTO suggestions for this URL. When the caller already
	1831	+ # warmed the transient cache (crawl-url job), those suggestions are
	1832	+ # handed in and NO second API call is made — the endpoint, rate
	1833	+ # limiter, and breaker are only hit once per URL.
	1834	+ # An empty array is a legitimate answer ("OTTO holds nothing for this
	1835	+ # URL") and must not be treated as a fetch failure; only a strict
	1836	+ # false from the fetcher is a failure.
	1837	+ if (is_array($prefetched_seo_data)) {
	1838	+ $seo_data = $prefetched_seo_data;
	1839	+ } else {
	1840	+ $failure_class = null;
	1841	+ $seo_data = metasync_fetch_otto_seo_data($route, $otto_uuid, $failure_class);
	1842	+ if ($seo_data === false) {
	1843	+ if ($failure_class === Metasync_Otto_Job_Status::OUTCOME_PERMANENT) {
	1844	+ # Auth/input-level rejection from the API — retrying the
	1845	+ # same request cannot help.
	1846	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
	1847	+ }
	1848	+ # Timeout / 5xx / 429 / empty body / bad JSON: retryable.
	1849	+ metasync_record_failed_action( 'metasync_process_seo_job' );
	1850	+ return Metasync_Otto_Job_Status::OUTCOME_RETRYABLE;
	1851	+ }
1126	1852	}
1127	1853
1128	1854	# Mark this URL as crawled by OTTO for SSR
1129	1855	# Extract domain and path from route
1130		- $parsed_url = parse_url($route);
	1856	+ $parsed_url = wp_parse_url($route);
1131	1857	$domain_with_scheme = ($parsed_url['scheme'] ?? 'https') . '://' . ($parsed_url['host'] ?? '');
1132	1858	$url_path = ($parsed_url['path'] ?? '/');
1133	1859
1134	1860	# Create crawl data structure
		@@ -1161,9 +1887,9 @@
1161	1887
1162	1888	# Try to find WooCommerce product by URL if url_to_postid failed
1163	1889	if ((!$post_id || $post_id <= 0) && strpos($route, '/product/') !== false && function_exists('wc_get_products')) {
1164	1890	# Extract product slug from URL
1165		- $product_slug = basename(parse_url($route, PHP_URL_PATH));
	1891	+ $product_slug = basename(wp_parse_url($route, PHP_URL_PATH));
1166	1892
1167	1893	# Try to get product by slug
1168	1894	$products = wc_get_products(array(
1169	1895	'name' => $product_slug,
		@@ -1194,17 +1920,16 @@
1194	1920	if (!$post_id || $post_id <= 0) {
1195	1921	# Check if this is a category page
1196	1922	if (strpos($route, '/category/') !== false) {
1197	1923	# Extract category slug from URL
1198		- $category_slug = basename(parse_url($route, PHP_URL_PATH));
	1924	+ $category_slug = basename(wp_parse_url($route, PHP_URL_PATH));
1199	1925	$category = get_category_by_slug($category_slug);
1200	1926
1201	1927	if ($category) {
1202	1928	# Check if category would return 404 before applying OTTO changes
1203	1929	if (metasync_would_term_return_404($category->term_id, 'category', $route)) {
1204		- error_log("MetaSync OTTO: Skipping SEO processing for category that would return 404: {$route} (Category ID: {$category->term_id})");
1205	1930	metasync_otto_auto_exclude_404_url($route);
1206		- return false;
	1931	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1207	1932	}
1208	1933
1209	1934	# Update comprehensive category SEO meta fields
1210	1935	$update_result = metasync_update_comprehensive_category_seo_fields($category->term_id, $seo_data);
		@@ -1267,8 +1992,12 @@
1267	1992	$title = 'Category Structured Data Update';
1268	1993	break;
1269	1994	}
1270	1995
	1996	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	1997	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	1998	+ }
	1999	+
1271	2000	if (!empty($title)) {
1272	2001	metasync_log_sync_history([
1273	2002	'title' => $title,
1274	2003	'source' => 'OTTO SEO',
		@@ -1284,12 +2013,12 @@
1284	2013	]);
1285	2014	}
1286	2015	}
1287	2016
1288		- return true;
	2017	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1289	2018	}
1290	2019
1291		- return false;
	2020	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1292	2021	}
1293	2022	}
1294	2023
1295	2024	# Check if this is a WooCommerce product category
		@@ -1294,17 +2023,16 @@
1294	2023
1295	2024	# Check if this is a WooCommerce product category
1296	2025	if (strpos($route, '/product-category/') !== false) {
1297	2026	# Extract product category slug from URL
1298		- $category_slug = basename(parse_url($route, PHP_URL_PATH));
	2027	+ $category_slug = basename(wp_parse_url($route, PHP_URL_PATH));
1299	2028	$term = get_term_by('slug', $category_slug, 'product_cat');
1300	2029
1301	2030	if ($term && !is_wp_error($term)) {
1302	2031	# Check if product category would return 404 before applying OTTO changes
1303	2032	if (metasync_would_term_return_404($term->term_id, 'product_cat', $route)) {
1304		- error_log("MetaSync OTTO: Skipping SEO processing for product category that would return 404: {$route} (Term ID: {$term->term_id})");
1305	2033	metasync_otto_auto_exclude_404_url($route);
1306		- return false;
	2034	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1307	2035	}
1308	2036
1309	2037	# Update comprehensive taxonomy SEO meta fields
1310	2038	$update_result = metasync_update_comprehensive_taxonomy_seo_fields($term->term_id, 'product_cat', $seo_data);
		@@ -1367,8 +2095,12 @@
1367	2095	$title = 'Product Category Structured Data Update';
1368	2096	break;
1369	2097	}
1370	2098
	2099	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2100	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2101	+ }
	2102	+
1371	2103	if (!empty($title)) {
1372	2104	metasync_log_sync_history([
1373	2105	'title' => $title,
1374	2106	'source' => 'OTTO SEO',
		@@ -1385,12 +2117,12 @@
1385	2117	]);
1386	2118	}
1387	2119	}
1388	2120
1389		- return true;
	2121	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1390	2122	}
1391	2123
1392		- return false;
	2124	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1393	2125	}
1394	2126	}
1395	2127
1396	2128	# Check if this is the home page (landing page)
		@@ -1411,11 +2143,10 @@
1411	2143
1412	2144	if ($home_page) {
1413	2145	# Check if home page would return 404 before applying OTTO changes
1414	2146	if (metasync_would_page_return_404($home_page->ID, $route)) {
1415		- error_log("MetaSync OTTO: Skipping SEO processing for home page that would return 404: {$route} (Post ID: {$home_page->ID}, Status: {$home_page->post_status})");
1416	2147	metasync_otto_auto_exclude_404_url($route);
1417		- return false;
	2148	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1418	2149	}
1419	2150
1420	2151	# Update comprehensive home page SEO meta fields
1421	2152	$update_result = metasync_update_comprehensive_seo_fields($home_page->ID, $seo_data);
		@@ -1481,8 +2212,12 @@
1481	2212	$title = 'Home Page Structured Data Update';
1482	2213	break;
1483	2214	}
1484	2215
	2216	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2217	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2218	+ }
	2219	+
1485	2220	if (!empty($title)) {
1486	2221	metasync_log_sync_history([
1487	2222	'title' => $title,
1488	2223	'source' => 'OTTO SEO',
		@@ -1497,12 +2232,12 @@
1497	2232	]);
1498	2233	}
1499	2234	}
1500	2235
1501		- return true;
	2236	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1502	2237	}
1503	2238
1504		- return false;
	2239	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1505	2240	}
1506	2241	}
1507	2242
1508	2243	# Check if this is the blog/posts page (page_for_posts)
		@@ -1513,11 +2248,10 @@
1513	2248
1514	2249	if ($posts_page && $route_clean === $posts_page_url) {
1515	2250	# Check if blog page would return 404
1516	2251	if (metasync_would_page_return_404($posts_page->ID, $route)) {
1517		- error_log("MetaSync OTTO: Skipping SEO processing for blog page that would return 404: {$route} (Post ID: {$posts_page->ID}, Status: {$posts_page->post_status})");
1518	2252	metasync_otto_auto_exclude_404_url($route);
1519		- return false;
	2253	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1520	2254	}
1521	2255
1522	2256	# Update comprehensive blog page SEO meta fields
1523	2257	$update_result = metasync_update_comprehensive_seo_fields($posts_page->ID, $seo_data);
		@@ -1583,8 +2317,12 @@
1583	2317	$title = 'Blog Page Structured Data Update';
1584	2318	break;
1585	2319	}
1586	2320
	2321	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2322	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2323	+ }
	2324	+
1587	2325	if (!empty($title)) {
1588	2326	metasync_log_sync_history([
1589	2327	'title' => $title,
1590	2328	'source' => 'OTTO SEO',
		@@ -1599,12 +2337,12 @@
1599	2337	]);
1600	2338	}
1601	2339	}
1602	2340
1603		- return true;
	2341	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1604	2342	}
1605	2343
1606		- return false;
	2344	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1607	2345	}
1608	2346	}
1609	2347
1610	2348	# URL didn't resolve to any supported entity (post, category, home page, blog page)
		@@ -1609,12 +2347,11 @@
1609	2347
1610	2348	# URL didn't resolve to any supported entity (post, category, home page, blog page)
1611	2349	# Treat as 404 and auto-exclude (e.g. deleted post, non-existent page)
1612	2350	if (!metasync_otto_is_url_available($route)) {
1613		- error_log("MetaSync OTTO: Skipping SEO processing for URL that would return 404 (no matching entity): {$route}");
1614	2351	metasync_otto_auto_exclude_404_url($route);
1615	2352	}
1616		- return false;
	2353	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1617	2354	}
1618	2355
1619	2356	# Verify this is actually a post, page, or WooCommerce product
1620	2357	$post = get_post($post_id);
		@@ -1623,16 +2360,15 @@
1623	2360	$supported_post_types = metasync_get_supported_post_types();
1624	2361
1625	2362	if (!$post || !in_array($post->post_type, $supported_post_types)) {
1626	2363	# Skip unsupported post types
1627		- return false;
	2364	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1628	2365	}
1629	2366
1630	2367	# Check if page would return 404 before applying OTTO changes
1631	2368	if (metasync_would_page_return_404($post_id, $route)) {
1632		- error_log("MetaSync OTTO: Skipping SEO processing for URL that would return 404: {$route} (Post ID: {$post_id}, Status: {$post->post_status})");
1633	2369	metasync_otto_auto_exclude_404_url($route);
1634		- return false;
	2370	+ return Metasync_Otto_Job_Status::OUTCOME_PERMANENT;
1635	2371	}
1636	2372
1637	2373	# Update comprehensive SEO meta fields
1638	2374	$update_result = metasync_update_comprehensive_seo_fields($post_id, $seo_data);
		@@ -1698,8 +2434,12 @@
1698	2434	$title = 'Structured Data Update';
1699	2435	break;
1700	2436	}
1701	2437
	2438	+ if (is_scalar($field_value) && trim((string) $field_value) === '') {
	2439	+ $title = str_replace(' Update (...)', ' Cleanup (previous value removed)', $title);
	2440	+ }
	2441	+
1702	2442	if (!empty($title)) {
1703	2443	metasync_log_sync_history([
1704	2444	'title' => $title,
1705	2445	'source' => 'OTTO SEO',
		@@ -1714,16 +2454,16 @@
1714	2454	]);
1715	2455	}
1716	2456	}
1717	2457
1718		- return true;
	2458	+ return Metasync_Otto_Job_Status::OUTCOME_SUCCESS;
1719	2459	}
1720	2460
1721		- return false;
	2461	+ return Metasync_Otto_Job_Status::OUTCOME_NO_CHANGE;
1722	2462
1723	2463	} catch (Exception $e) {
1724	2464	metasync_record_failed_action( 'metasync_process_seo_job' );
1725		- return false;
	2465	+ return Metasync_Otto_Job_Status::OUTCOME_RETRYABLE;
1726	2466	} finally {
1727	2467	# Release the concurrency lock only if we actually acquired it.
1728	2468	# Early returns (CPU deferral, lock contention) must NOT delete a lock
1729	2469	# that another process may be holding.
		@@ -1744,11 +2484,10 @@
1744	2484
1745	2485	// Minimal duplicate prevention within short time window
1746	2486	if (!empty($data['title']) && !empty($data['source'])) {
1747	2487	global $wpdb;
1748		- $table = $wpdb->prefix . Metasync_Sync_History_Database::$table_name;
1749		- $recent = $wpdb->get_var($wpdb->prepare(
1750		- "SELECT COUNT(*) FROM `$table` WHERE title = %s AND source = %s AND created_at >= %s",
	2488	+ $recent = $wpdb->get_var($wpdb->prepare( // phpcs:ignore WordPress.DB.DirectDatabaseQuery -- sync-history duplicate guard over the plugin custom table — write-path freshness required
	2489	+ "SELECT COUNT(*) FROM `{$wpdb->prefix}metasync_sync_history` WHERE title = %s AND source = %s AND created_at >= %s",
1751	2490	$data['title'],
1752	2491	$data['source'],
1753	2492	gmdate('Y-m-d H:i:s', time() - 60)
1754	2493	));
		@@ -1759,9 +2498,9 @@
1759	2498
1760	2499	$sync_db->add($data);
1761	2500
1762	2501	} catch (Exception $e) {
1763		- error_log("MetaSync: Failed to log sync history: " . $e->getMessage());
	2502	+ 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
1764	2503	}
1765	2504	}
1766	2505
1767	2506	/**
		@@ -1837,9 +2576,9 @@
1837	2576	$db = new Metasync_Redirection_Database();
1838	2577	$redirect = new Metasync_Redirection($db);
1839	2578	return $redirect->resolve_url_to_final_destination($url, 10);
1840	2579	} catch (Exception $e) {
1841		- error_log('MetaSync OTTO: Redirect resolution failed for ' . $url . ' - ' . $e->getMessage());
	2580	+ 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
1842	2581	return $url;
1843	2582	}
1844	2583	}
1845	2584
		@@ -1870,9 +2609,9 @@
1870	2609	'status' => 'active',
1871	2610	'auto_excluded' => 1,
1872	2611	]);
1873	2612	} catch (Exception $e) {
1874		- error_log('MetaSync OTTO: Failed to auto-exclude 404 URL: ' . $url . ' - ' . $e->getMessage());
	2613	+ 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
1875	2614	return false;
1876	2615	}
1877	2616	}
1878	2617
		@@ -1892,14 +2631,13 @@
1892	2631	try {
1893	2632	require_once plugin_dir_path(__FILE__) . 'class-metasync-otto-excluded-urls-database.php';
1894	2633	$db = new Metasync_Otto_Excluded_URLs_Database();
1895	2634	global $wpdb;
1896		- $table = $wpdb->prefix . Metasync_Otto_Excluded_URLs_Database::$table_name;
1897	2635	# Normalize the same way is_url_excluded() does
1898	2636	$url_normalized = rtrim(trim($url), '/');
1899		- $records = $wpdb->get_results(
	2637	+ $records = $wpdb->get_results( // phpcs:ignore WordPress.DB.DirectDatabaseQuery -- webhook-path cleanup of auto-exclusions over the plugin custom table
1900	2638	$wpdb->prepare(
1901		- "SELECT id FROM `{$table}` WHERE url_pattern = %s AND auto_excluded = 1 AND status = 'active'",
	2639	+ "SELECT id FROM `{$wpdb->prefix}metasync_otto_excluded_urls` WHERE url_pattern = %s AND auto_excluded = 1 AND status = 'active'",
1902	2640	$url_normalized
1903	2641	)
1904	2642	);
1905	2643	if (!empty($records)) {
		@@ -1929,24 +2667,26 @@
1929	2667	}
1930	2668	try {
1931	2669	require_once plugin_dir_path(__FILE__) . 'class-metasync-otto-excluded-urls-database.php';
1932	2670	global $wpdb;
1933		- $table = $wpdb->prefix . Metasync_Otto_Excluded_URLs_Database::$table_name;
1934	2671	$url_normalized = rtrim(trim($url), '/');
1935	2672
1936	2673	$records = get_transient(METASYNC_OTTO_EXCLUDED_TRANSIENT_KEY);
1937	2674
1938	2675	if ($records === false) {
1939		- // phpcs:ignore WordPress.DB.PreparedSQL.NotPrepared -- no user input, table name from $wpdb->prefix
1940		- $records = $wpdb->get_results(
1941		- "SELECT url_pattern, pattern_type FROM `{$table}` WHERE status = 'active' AND (auto_excluded = 0 OR auto_excluded IS NULL) ORDER BY created_at DESC"
	2676	+ $records = $wpdb->get_results( // phpcs:ignore WordPress.DB.DirectDatabaseQuery -- render-path read IS cached via get/set_transient with invalidation on every write — wp_cache_* is not used because the payload must survive object-cache-only setups
	2677	+ "SELECT url_pattern, pattern_type FROM `{$wpdb->prefix}metasync_otto_excluded_urls` WHERE status = 'active' AND (auto_excluded = 0 OR auto_excluded IS NULL) ORDER BY created_at DESC"
1942	2678	);
1943	2679
1944	2680	// Graceful recovery: auto_excluded column missing on pre-v2.7.4 installs.
1945		- // Run ALTER TABLE to add it and treat URL as not excluded so OTTO continues rendering.
	2681	+ // Schema changes belong to the upgrade routines, never the render path:
	2682	+ // degrade to "not excluded" so OTTO keeps rendering, and schedule a
	2683	+ // one-shot background repair that runs the same column migration off
	2684	+ // the visitor request (handled by metasync_repair_otto_excluded_schema).
1946	2685	if ($wpdb->last_error && strpos($wpdb->last_error, 'auto_excluded') !== false) {
1947		- // phpcs:ignore WordPress.DB.PreparedSQL.NotPrepared
1948		- $wpdb->query("ALTER TABLE `{$table}` ADD COLUMN `auto_excluded` TINYINT(1) NOT NULL DEFAULT 0");
	2686	+ if (false === wp_next_scheduled('metasync_repair_otto_excluded_schema')) {
	2687	+ wp_schedule_single_event(time() + MINUTE_IN_SECONDS, 'metasync_repair_otto_excluded_schema');
	2688	+ }
1949	2689	return false;
1950	2690	}
1951	2691
1952	2692	set_transient(METASYNC_OTTO_EXCLUDED_TRANSIENT_KEY, $records ?: [], METASYNC_OTTO_EXCLUDED_TRANSIENT_TTL);
		@@ -1956,27 +2696,16 @@
1956	2696	return false;
1957	2697	}
1958	2698
1959	2699	foreach ($records as $excluded) {
1960		- $pattern = rtrim(trim($excluded->url_pattern), '/');
1961		- $pattern_type = $excluded->pattern_type;
1962		-
1963		- switch ($pattern_type) {
1964		- case 'exact':
1965		- if ($url_normalized === $pattern) {
1966		- return true;
1967		- }
1968		- break;
1969		- case 'contain':
1970		- if (strpos($url_normalized, $pattern) !== false) {
1971		- return true;
1972		- }
1973		- break;
1974		- case 'start':
1975		- if (strpos($url_normalized, $pattern) === 0) {
1976		- return true;
1977		- }
1978		- break;
	2700	+ # Delegate to the shared matcher so this gate supports every pattern type the
	2701	+ # add handler accepts (exact|contain|start|end|regex). Keeping a second switch
	2702	+ # here is what let `end` and `regex` exclusions be silently ignored on the
	2703	+ # visitor render path while still applying on the queue/SEO-write path.
	2704	+ # The raw pattern is passed through - the matcher normalizes it per type,
	2705	+ # because a regex must keep the trailing slash that closes its delimiters.
	2706	+ if (Metasync_Otto_Excluded_URLs_Database::pattern_matches($url_normalized, $excluded->url_pattern, $excluded->pattern_type)) {
	2707	+ return true;
1979	2708	}
1980	2709	}
1981	2710	return false;
1982	2711	} catch (Exception $e) {
		@@ -2033,9 +2762,9 @@
2033	2762	}
2034	2763
2035	2764	# WooCommerce product by slug
2036	2765	if ((!$post_id || $post_id <= 0) && strpos($route, '/product/') !== false && function_exists('wc_get_products')) {
2037		- $product_slug = basename(parse_url($route, PHP_URL_PATH));
	2766	+ $product_slug = basename(wp_parse_url($route, PHP_URL_PATH));
2038	2767	$products = wc_get_products(array('name' => $product_slug, 'limit' => 1, 'status' => 'publish'));
2039	2768	if (!empty($products)) {
2040	2769	$post_id = $products[0]->get_id();
2041	2770	} else {
		@@ -2062,9 +2791,9 @@
2062	2791	}
2063	2792
2064	2793	# Category
2065	2794	if (strpos($route, '/category/') !== false) {
2066		- $category_slug = basename(parse_url($route, PHP_URL_PATH));
	2795	+ $category_slug = basename(wp_parse_url($route, PHP_URL_PATH));
2067	2796	$category = get_category_by_slug($category_slug);
2068	2797	if ($category) {
2069	2798	return !metasync_would_term_return_404($category->term_id, 'category', $route);
2070	2799	}
		@@ -2071,9 +2800,9 @@
2071	2800	}
2072	2801
2073	2802	# WooCommerce product category
2074	2803	if (strpos($route, '/product-category/') !== false) {
2075		- $category_slug = basename(parse_url($route, PHP_URL_PATH));
	2804	+ $category_slug = basename(wp_parse_url($route, PHP_URL_PATH));
2076	2805	$term = get_term_by('slug', $category_slug, 'product_cat');
2077	2806	if ($term && !is_wp_error($term)) {
2078	2807	return !metasync_would_term_return_404($term->term_id, 'product_cat', $route);
2079	2808	}
		@@ -2115,9 +2844,9 @@
2115	2844
2116	2845	$removed = 0;
2117	2846	$marked_permanent = 0;
2118	2847	$thirty_days_ago = strtotime('-30 days');
2119		- $next_recheck = date('Y-m-d H:i:s', current_time('timestamp') + 7 * DAY_IN_SECONDS);
	2848	+ $next_recheck = gmdate('Y-m-d H:i:s', current_time('timestamp') + 7 * DAY_IN_SECONDS);
2120	2849
2121	2850	foreach ($records as $record) {
2122	2851	$url = trim($record->url_pattern);
2123	2852	if (empty($url)) {
		@@ -2138,16 +2867,10 @@
2138	2867	}
2139	2868	}
2140	2869	}
2141	2870
2142		- if ($removed > 0) {
2143		- error_log("MetaSync OTTO: Recheck 404 exclusions - removed {$removed} URL(s) that are now available");
2144		- }
2145		- if ($marked_permanent > 0) {
2146		- error_log("MetaSync OTTO: Recheck 404 exclusions - marked {$marked_permanent} URL(s) as permanent (still 404 after 30 days)");
2147		- }
2148	2871	} catch (Exception $e) {
2149		- error_log('MetaSync OTTO: Recheck 404 exclusions failed - ' . $e->getMessage());
	2872	+ 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
2150	2873	}
2151	2874	}
2152	2875
2153	2876	add_action('metasync_otto_recheck_404_exclusions', 'metasync_otto_recheck_404_exclusions');
		@@ -2152,8 +2875,24 @@
2152	2875
2153	2876	add_action('metasync_otto_recheck_404_exclusions', 'metasync_otto_recheck_404_exclusions');
2154	2877
2155	2878	/**
	2879	+ * Background repair of the metasync_otto_excluded_urls schema (cron).
	2880	+ *
	2881	+ * Scheduled by metasync_is_otto_url_manually_excluded() when a render-path
	2882	+ * read fails because the auto_excluded column is missing (pre-2.7.4 install
	2883	+ * whose version migration did not run). Runs the DB layer's self-healing
	2884	+ * column migration off the visitor request instead of ALTERing inline.
	2885	+ */
	2886	+function metasync_repair_otto_excluded_schema()
	2887	+{
	2888	+ require_once plugin_dir_path(__FILE__) . 'class-metasync-otto-excluded-urls-database.php';
	2889	+ Metasync_Otto_Excluded_URLs_Database::ensure_schema();
	2890	+}
	2891	+
	2892	+add_action('metasync_repair_otto_excluded_schema', 'metasync_repair_otto_excluded_schema');
	2893	+
	2894	+/**
2156	2895	* Check if a post/page would return 404 without making HTTP request
2157	2896	* Uses WordPress database checks for fast validation
2158	2897	*
2159	2898	* @param int $post_id WordPress post ID
		@@ -2261,5 +3000,88 @@
2261	3000	# Check if this post has Brizy metadata
2262	3001	if (get_post_meta($post_id, 'brizy_post_uid', true)) {
2263	3002	delete_transient('metasync_has_brizy_posts');
2264	3003	}
2265		-}, 10, 1);
	3004	+}, 10, 1);
	3005	+
	3006	+/**
	3007	+ * Identify infrastructure agents that must never hit the render throttle.
	3008	+ *
	3009	+ * Two classes of agent are covered:
	3010	+ *
	3011	+ * - Page-cache warmers. A host preloader (WP Cloud PageCacheBot, WP Rocket preload,
	3012	+ * LiteSpeed, SG Optimizer, ...) exists purely to populate the page cache. Throttling
	3013	+ * one returns un-OTTO'd HTML *and* defines DONOTCACHEPAGE, so the cache can never be
	3014	+ * filled and every subsequent human visitor pays a full uncached render. Measured on
	3015	+ * the reporting site: desktop cache MISS on 17/17 requests at ~2.8s TTFB, versus
	3016	+ * ~0.56s with OTTO disabled.
	3017	+ *
	3018	+ * - Synthetic performance auditors (Lighthouse / PageSpeed Insights / GTmetrix).
	3019	+ * These are what customers measure with, so they must receive the same fully
	3020	+ * rendered OTTO output a real visitor gets. Currently `/lighthouse/i` and
	3021	+ * `/pagespeed/i` are generic-bot patterns, so a second PSI run inside the 5-minute
	3022	+ * window is served the throttled, un-OTTO'd, uncacheable page.
	3023	+ *
	3024	+ * Matching is done on the raw user-agent, deliberately independent of the bot-name
	3025	+ * categorisation in Metasync_Otto_Bot_Detector.
	3026	+ *
	3027	+ * @param mixed $detection Result of Metasync_Otto_Bot_Detector::detect(), or malformed input.
	3028	+ * @return bool True when this request must bypass the render throttle.
	3029	+ */
	3030	+function metasync_otto_is_unthrottled_infrastructure_agent( $detection = array() ) {
	3031	+
	3032	+ $ua = '';
	3033	+ if ( is_array( $detection ) && ! empty( $detection['user_agent'] ) ) {
	3034	+ $ua = (string) $detection['user_agent'];
	3035	+ } elseif ( isset( $_SERVER['HTTP_USER_AGENT'] ) ) {
	3036	+ $ua = (string) $_SERVER['HTTP_USER_AGENT'];
	3037	+ }
	3038	+
	3039	+ $needles = array(
	3040	+ // Page-cache preloaders.
	3041	+ 'pagecachebot', // WP Cloud / Automattic (PageCacheBotDesktop|Mobile)
	3042	+ 'wp rocket', // WP Rocket preload
	3043	+ 'wprocket',
	3044	+ 'lscache_runner', // LiteSpeed Cache crawler
	3045	+ 'sg-optimizer', // SiteGround
	3046	+ 'sgoptimizer',
	3047	+ 'nitropack',
	3048	+ 'cache-warmer',
	3049	+ 'cachewarmer',
	3050	+ 'cache warmer',
	3051	+ // Synthetic performance auditors.
	3052	+ 'lighthouse',
	3053	+ 'pagespeed',
	3054	+ 'gtmetrix',
	3055	+ 'pingdom',
	3056	+ 'webpagetest',
	3057	+ );
	3058	+
	3059	+ $is_infra = false;
	3060	+
	3061	+ if ( '' !== $ua ) {
	3062	+ $ua_lower = strtolower( $ua );
	3063	+ foreach ( $needles as $needle ) {
	3064	+ if ( false !== strpos( $ua_lower, $needle ) ) {
	3065	+ $is_infra = true;
	3066	+ break;
	3067	+ }
	3068	+ }
	3069	+ }
	3070	+
	3071	+ // WP Cloud's warmer also tags its requests with a query marker.
	3072	+ if ( ! $is_infra && isset( $_GET['x-cache-engine'] ) ) { // phpcs:ignore WordPress.Security.NonceVerification.Recommended -- read-only flag for render throttling
	3073	+ $is_infra = true;
	3074	+ }
	3075	+
	3076	+ /**
	3077	+ * Allow a site or host to declare additional preload agents.
	3078	+ *
	3079	+ * @param bool $is_infra Whether this request bypasses the OTTO render throttle.
	3080	+ * @param string $ua The request user-agent.
	3081	+ */
	3082	+ if ( function_exists( 'apply_filters' ) ) {
	3083	+ $is_infra = (bool) apply_filters( 'metasync_otto_unthrottled_agent', $is_infra, $ua );
	3084	+ }
	3085	+
	3086	+ return $is_infra;
	3087	+}