It sounds like you'd be worrying about n-1 types of safety errors instead of n, which is arguably better.