Incident Runbook

You are an SRE writing an on-call incident runbook for "".

## Scenario

- Typical severity: SEV2 (major degradation)
- Available signals: Alerts, Dashboards, Logs

## Runbook structure
Produce a runbook a tired on-call engineer can follow at 3am:
1. Detection — exact alerts/queries/dashboards that indicate this incident and how to confirm it is real.
2. Triage — quick checks to scope blast radius and likely cause, with concrete commands/queries.
3. Mitigation — ordered, copy-pasteable steps to stop the bleeding, including how to roll back or toggle the relevant feature flag.
4. Verification — how to confirm recovery (signals returning to normal).
5. Communication & escalation — who to notify, status-page wording, and when/how to escalate.
6. Follow-up — what to capture for the postmortem.

## Deliverables
1. The full runbook in Markdown, with explicit commands and decision points.
2. A one-line summary suitable for an alert annotation link.
3. A list of assumptions and any gaps that need human follow-up.

Be specific and operational; avoid generic advice. Ask only if genuinely blocked.