Scheduler and worker loops survive a database restart

index-scheduler's loop ended when Postgres restarted (the active-groups
query was outside its try), leaving live indexing stopped for ~3 hours
while the process stayed up. Each step's error is now caught and retried
next round; worker:spots and worker:deferred wait 30 s and claim again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

commands/IndexScheduler.ts

@@ -117,11 +117,18 @@ export default class IndexScheduler extends BaseCommand {
       }
     }
 
-    // Run immediately, then wait a full interval after each run so runs never overlap.
+    // Run immediately, then wait a full interval after each run so runs never overlap. A failed
+    // round (e.g. Postgres restarting) is logged and the next one tries again: an error that escaped
+    // here once ended the loop while the process stayed up, and live indexing stopped for ~3 hours
+    // (LOG.md 2026-10-07 (3)).
     while (!Shutdown.requested) {
-      await schedule()
-      await backfill()
-      await split()
+      for (const step of [schedule, backfill, split]) {
+        try {
+          await step()
+        } catch (err: any) {
+          this.logger.error(`Scheduler round failed, retrying next round: ${err.message}`)
+        }
+      }
       await Shutdown.sleep(indexerConfig.intervalMs)
     }
     await this.pool.shutdown()