Runbook: OOMKilled / OutOfMemoryError

Symptom: container exits 137, OOMKilled in pod events, heap-exhausted errors in application logs.

Triage:

1. Confirm the kill is the kubelet OOM killer (exit 137) and not an application-level limit.
2. Open the deployment manifest under infra/k8s/<service>-deploy.yaml and check spec.template.spec.containers[].resources.limits.memory.
3. If the limit is below the service's documented working set (checkout-svc needs ~400Mi after the pricing cache warms), raise the limit. 512Mi is the standard next tier.
4. Keep requests.memory at or below the new limit.

Fix: open a PR against the infra repo with the corrected limit. Do not hot-patch the live deployment.