For now, let's just bite the bullet and require 100% patch coverage from here on out. If we hit a really nasty case where it really is a significant investment of effort, then maybe we can revisit this discussion and explore a better way to express this exception without losing information about test coverage completely.