The waitgroup counter goes up before we even try to start the service. So, if the start fails, the counter doesn't go back down. This messes up the graceful shutdown of cgr-engine.
