Observability & Incident Response

• Quiet API p95 latency. User-facing latency above the teaching SLO window.
• Quiet Checkout error ratio. 5xx on /checkout. Not explained by host CPU.
• Quiet Replica CPU saturation. Ready pods above the utilization target; Pending replacements.

No firing alerts.

1. T+00:12 GET /healthz 200 12ms · tr-100
2. T+04:00 GET /checkout 200 74ms · tr-101