Incident-response demonstration boundary

The current AKS cluster has been deleted. The repository contains a local kind alert fire/resolve proof, not production incident-response SLOs or contracted P0/P1 response times.

For a local interview walkthrough, use the recorded alert cycle: the metric source became unavailable, LocalPlatformMetricSourceDown fired, the source recovered, and the alert resolved. State the environment, timestamps and what was not measured. The separate self-heal record proves two Argo-owned resource recovery cases on kind.

For a real incident, first identify the exact cluster context and affected service, preserve logs/metrics and timeline, assign an incident owner, and select the owning workload's rollback or repair runbook. A manual kubectl rollout undo against an Argo-managed resource can be overwritten by GitOps; the reviewed Git source must express the intended rollback. Azure commands in the historical teardown runbook are not a routine incident demo.