fix: improve sync resilience and make health staleness configurable

src/config/configLoader.ts

@@ -37,6 +37,9 @@ function loadConfig(): Config {
       port: process.env.HEALTH_PORT
         ? parseInt(process.env.HEALTH_PORT, 10)
         : undefined,
+      maxStaleMs: process.env.HEALTH_MAX_STALE_MS
+        ? parseInt(process.env.HEALTH_MAX_STALE_MS, 10)
+        : undefined,
     },
   };
 

src/config/configSchema.ts

@@ -35,6 +35,11 @@ export const healthConfigSchema = z.object({
   enabled: z.boolean().default(true),
   host: z.string().default('::'),
   port: z.number().int().positive().default(3000),
+  maxStaleMs: z
+    .number()
+    .int()
+    .positive()
+    .default(10 * 60 * 1000), // 10 minutes
 });
 
 export const configSchema = z.object({

src/health.ts

@@ -37,7 +37,10 @@ export function createHealthRequestHandler({
     }
 
     const postgresStatus = await performPostgresCheck(pool, logger);
-    const synchronizerStatus = evaluateSynchronizer(synchronizer);
+    const synchronizerStatus = evaluateSynchronizer(
+      synchronizer,
+      config.health.maxStaleMs,
+    );
 
     const isHealthy =
       postgresStatus.status === 'ok' && synchronizerStatus.status === 'ok';
@@ -88,6 +91,7 @@ export async function performPostgresCheck(
 
 export function evaluateSynchronizer(
   synchronizer: Synchronizer,
+  maxStaleMs: number,
 ): HealthComponentStatus {
   const metrics = synchronizer.getMetrics();
   if (!metrics.lastSyncTime) {
@@ -110,7 +114,6 @@ export function evaluateSynchronizer(
   }
 
   const now = Date.now();
-  const maxStaleMs = 10 * 60 * 1000; // 10 minutes
   const timeSinceLastSync = now - lastSyncTimestamp;
 
   if (timeSinceLastSync > maxStaleMs) {

src/synchronizer/meilisearch/createMeiliSearchSynchronizer.ts

@@ -88,23 +88,6 @@ export function createMeiliSearchSynchronizer(
           stack: error instanceof Error ? error.stack : undefined,
         },
       });
-      // Stop change detection and reset state on sync failure.
-      isRunning = false;
-      await changeDetection.stop().catch(stopError => {
-        logger.warn(
-          'Error while stopping change detection after sync failure:',
-          {
-            metadata: {
-              error:
-                stopError instanceof Error
-                  ? stopError.message
-                  : String(stopError),
-            },
-          },
-        );
-      });
-
-      throw error;
     }
   };