Bug Fix
Brief downtime during Kubernetes password rotation
Issue
During a coordinated PostgreSQL password change, the Kubernetes pods were restarted. New pods initially failed to connect due to a secret encoding issue with the ! character in the password, causing a brief CrashLoopBackOff state (~2-3 minutes of degraded service).
Resolution
Secret was recreated using proper base64 encoding instead of --from-literal. All pods recovered and are now running normally.
Bonus: The pod restart picked up the recent Sidekiq queue fix, resulting in significantly improved forum responsiveness.
Impact
~3 minutes of intermittent errors. No data loss. Performance is now noticeably faster due to proper background job processing.