A running process may still be unable to serve useful work. A healthcheck runs a small diagnostic command and records whether it succeeds. Use a check that reflects the application boundary you need, rather than assuming process existence proves readiness. A failed healthcheck can mark a container unhealthy while its main process continues running. Restart behavior depends on the surrounding platform; the health label alone doesn't guarantee a restart. In these local tasks, a deterministic command result makes that distinction visible. You'll configure successful checks and one deliberately failing check, then inspect health without deleting the evidence or stopping the application.