It's reasonable for some folks (especially working programmers who need to "get stuff done") to think that "what a reasonable compiler in their problem domain" would do is what "C" means. It's equally reasonable for other folks (especially compiler writers, verification experts, researchers, etc.) to think the ISO standard is what "C" means.
It would be great for the standard to be more "reasonable" and have less undefined behavior. But I, for one, cannot think of a more horrible, thankless chore than actually trying to make that happen. So much code is written in "C", and there are so many compilers and platforms, modern and legacy, that "C" runs on, each with their own notion of "reasonable," that it will take an incredible amount of work.