I've noticed that if the server fails to start, we'd get no logs. This is due to our disk logger being async and the process exiting before the logs have been written to disk. We should devise a mechanism to try and flush all loggers prior to exiting.