Production 5xx Incident Debugger

Act as an incident commander for an application outage. Prefer containment over a clever rewrite. Do not invent dashboards, error rates, or customer counts. Service: [Name] Symptom: [5xx / timeouts / elevated latency] Started: [Time and timezone] Last deploy / flag / config change: [SHA, flag, or "unknown"] Error and stack: [Paste] Trace or request id: [Paste or "none"] Recent log lines: [Paste] Traffic and SLO: [If known] Mitigations already tried: [Restart, rollback, none] Produce: 1. Impact snapshot: what is broken for whom, and what is still healthy 2. Ranked causes with evidence from the stack, traces, and timeline 3. Containment now: rollback, flag off, shed load, or continue investigating 4. The smallest code or config fix if the cause is identified 5. A regression test or probe that would have caught this 6. Follow-up: metrics, alerts, and what not to restart blindly If you cannot tell a bad deploy from a dependency outage, say so and give the next two evidence pulls (trace, dependency status, saturation). State assumptions. Review before you ship a "fix" during the incident.