Noted, but the point of the drill is precisely to uncover these failure modes and attempt to fix them. e.g. you might have automated runbooks to fix the problem rather than access the DB directly. You might have frequent backups and processes to easily restore from backups in case of database wipes.