Ship production optimization M1-M12.5A: queues, metrics, HTTP observability, and vector search reliability.

Keep similar-ai from tripping the global circuit on a lone URL 502, clamp Qdrant search to 100, and add Server-Timing plus slow-request logging. Studio shared props, Academy S3 exists caching, heat chunking, and Redis/scheduler hygiene stay in this rollout.

app/Jobs/IndexArtworkJob.php

@@ -6,6 +6,7 @@ namespace App\Jobs;
 
 use App\Models\Artwork;
 use Illuminate\Bus\Queueable;
+use Illuminate\Contracts\Queue\ShouldBeUniqueUntilProcessing;
 use Illuminate\Contracts\Queue\ShouldQueue;
 use Illuminate\Foundation\Bus\Dispatchable;
 use Illuminate\Queue\InteractsWithQueue;
@@ -21,16 +22,29 @@ use Meilisearch\Client as MeilisearchClient;
  * after_commit double-dispatch problem and ensures the document lands
  * in the index within this job's execution, with no extra queue hop.
  */
-class IndexArtworkJob implements ShouldQueue
+class IndexArtworkJob implements ShouldQueue, ShouldBeUniqueUntilProcessing
 {
     use Dispatchable, InteractsWithQueue, Queueable, SerializesModels;
 
     public int $tries   = 3;
     public int $timeout = 60;
 
+    /**
+     * Unique while queued so Meilisearch isn't flooded. handle() loads the
+     * artwork from DB, so a dropped duplicate still indexes current state.
+     * UniqueUntilProcessing: a change after the worker starts can enqueue again.
+     */
+    public int $uniqueFor = 120;
+
     public function __construct(public readonly int $artworkId)
     {
         $this->afterCommit = true;
+        $this->onQueue((string) config('scout.queue.queue', 'search'));
+    }
+
+    public function uniqueId(): string
+    {
+        return (string) $this->artworkId;
     }
 
     public function handle(MeilisearchClient $client): void