API latency regression

Overview

Use this when dashboards show elevated p95/p99 or 5xx without a clear deploy. Validate scope before paging dependent teams.

Customer-facing latency or error-rate spike on edge/API tier.

Step checks

1. 1

   Confirm blast radius

   Scope: single region vs global; guest vs auth traffic.

2. 2

   Correlate deploys & flags

   Last 2h deploys, feature flags, autoscaling events.

3. 3

   Stabilize

   Traffic shift, rollback, or capacity bump with owner sign-off.

4. 4

   Communicate

   Status page + internal channel with ETA for next update.