I don't know what to expect tbh, and I'm also even fine with the current behaviour :)
AFAIK the C committee is on the right track though. Start with collecting and classifying all mentions of UB in the standard (IIRC it's somewhere around 200 items), then decide what to do with each of those. A lot of the UB items are completely silly ancient artifacts which can easily be clarified (like the "no newline at the end of a source file").
Then there's probably quite a few items that can become implementation defined behaviour instead.
And only then decide what should happen to the remaining UB items that are actually important for optimization, maybe require a new 'checked mode' which turns those cases into runtime errors.
Also I think the "nasal demons" problem needs to tackled somehow (e.g. the thing that a compiler is allowed to do anything in the presence of UB) - obviously this is the trickiest part, and in reality there's hardly any C program which doesn't contain UB.