This is why production services should have health checks that actually verify whether the service is doing its job in the system, not whether it's merely running the process. A process can be completely broken while still technically running, and vice versa (at least with janky ps-parsing scripts).