This can be used in k8s and wherever else a liveness check is required.
This deviates a bit from our other liveliness checks (and I think the standard) in that instead of actually looking at the process inside the container, it asks the DB. I think I prefer this over adding some port on which the daemon health can be checked. This keeps us with a single source of truth for daemon liveliness. If the db goes down then this container will be considered unhealthy, but the container would be crashing anyway. I might be overlooking some pitfall though, so I'd appreciate feedback here.