The status.midos.house infrastructure helps avoid restarting the server during a race. Regardless, crashes and emergency restarts do happen, and it would be nice to be able to give a better user experience in that event. Keep a backup in the database of as much of the full race handler state as is reasonable, restoring from it when starting up a new race handler, and cleaning up rows that don't correspond to ongoing races.