Supabase Pooler Connection Timeout Causes Full Production Outage In Us-East-1

Supabase managed connection pooler (pgBouncer/Supavisor) in us-east-1 did not fully recover after maintenance, leaving stale or unhealthy connections that cause new queries to time out. The database itself is healthy as shown by direct connections and idle pg_stat_activity, but the pooler is unable to establish or reuse connections reliably.

1. Send any data query via Supabase REST API or client libraries to project wwxjopfybrfomtealpjc. 2. Observe request hangs indefinitely or returns 503 with 'Connection terminated due to connection timeout'. 3. Test direct database connection using the direct connection string; approximately 1 in 3 attempts succeed. 4. On successful direct connection, run pg_stat_activity and see mostly idle backends with no application queries, confirming database health. 5. Check Supabase status page for us-east-1; incidents are marked resolved but the issue persists.

Using direct connections bypasses connection pooling, which may exhaust database connection limits under high concurrency. This workaround should only be used until the pooler is confirmed healthy. After Supabase resolves the incident, revert DATABASE_URL to the pooler URL and restart the application. In multi-threaded or distributed environments, ensure the environment variable change is propagated atomically. Test with a small subset of traffic before full rollout. Rollback: restore original DATABASE_URL and restart services.