Arvados

For all services:

If the config file on disk changes, it checks that the config file is valid (config-check), then (if it is valid) restarts the service.

If the config file is not valid, that is reported via health check (possibly already implemented?)

Ideally, restarting the service means allowing existing in-flight requests to complete, while new requests use the config.

When running under "arvados-server boot", this is currently handled by a supervisor process. When not running under "arvados-server boot" services need to monitor the config file & be able to reload/restart themselves.