Off-infrastructure dead-man probe for grandflipout.com.
Every watchdog inside the stack runs on the stack it monitors — a total outage would take the alarms down with it. This tiny scheduled workflow runs on GitHub-hosted runners instead: every ~10 minutes it probes the app origin's health endpoints, retries three times a minute apart, and on sustained failure posts one alert to a private ops channel.
No secrets in code (the alert webhook is a repo secret), no data collected, nothing to deploy. GitHub cron is best-effort — treat this as a dead-man layer, not a latency monitor.