Fix: systemd: Ensure the retry logic is also exercised for the 'service' class

Also use up all of the configured timeout
Log when we give up due to timeouts