- Container startup loop caused by insufficient PostgreSQL privileges in the startup permissions script.
- Wrapped the `ALTER ROLE ... CREATEROLE` in a safe DO block with exception handling so missing privilege is logged as a NOTICE and startup continues.
- No feature changes; service now starts normally without elevated DB privileges.
- _Files: `backend/fix_permissions.sql`_
9f123776e8