Last edited 4 days ago by Barbara Liu.
browse-service health check fails, page the engineering on-call (rotation: Eng-Rota). Don't restart the cluster — bounce the affected pod only.
1. Cache backpressure: Redis OOM warnings → drain queue, scale Redis up.
2. API drift: 5xx spike after deploy → compare with last known good ref.
3. Stale tokens: 401s spike → rotate the IDP secret in Vault.