AWS is green. The product is not. I stay until it is.

AWS status isn’t the outage. Failed failover, a deploy without rollback, the same page again — users failing while the dashboards stay polite.

A deploy without rollback has an answer I reach for: a canary on a Lambda alias. Traffic shifts back. The source layer only moves forward.

For a director, VP, or principal in an active or just-landed incident, or a failover that’s never been proven. Not a Well-Architected slide, not “Amazon will handle it,” not a review with nobody in the pager.

If it’s down: email with LIVE in the subject. First reply is whether I can take it, not incident command from this website. If we start, I join your channel and pager, with access and NDA in the thread. After it holds we harden — a symptom cleared with the cause still in the path is how you get the same runbook, the same people, and the same page next quarter. The same incident should not need you twice.