Bounds checking is not zero cost in all cases. So long as you can demonstrate even one example of a case where the bounds checking isn't zero cost, and yet you can prove that the bounds check isn't needed there will be push back. C++ is about don't pay for what you don't use, and if you don't need the bounds check you shouldn't have to pay for it. You cannot have high performance programing if there are various 'little' if statements added all over.
Remember that when proving a program you often have more information than the C++ compiler does. The compiler never see the whole program, only the part you are building now. Thus it is easy to show the above just by having the calling function and callie function in different translations units.
If high performance isn't one of your goals, then why are you writing C++ at all?
Note, languages like rust do should there is opportunity to do better than C++ without a runtime cost. However they do not provide perfect protection, either there are cases where they cannot do the bounds check, or there are times where they pay the price of a not needed bounds check. All such languages prove is C++ can do better in some cases.