migrationutil::submitPendingDeletions can throw (e.g., if PersistentTaskStore::forEach throws, which can happen at least if a range deletion doc cannot be parsed), which will cause resubmitRangeDeletionsOnStepUp to terminate the server since it would be an unhandled exception.
At the very least, we should add a try/catch in resubmitRangeDeletionsOnStepUp.