I would be okay losing it if the process died and I started a new one; I’d check against persistent state to see if I needed to start a new timeout or immediately do the work that should have been done while the process was dead.