This is addressed in the footnotes. casting is not the core of the type safety. Read the whole article.
I also don’t agree it’s “squeamish” to be wary of aliasing analysis going wrong. It’s not a clean abstraction and can hide subtle bugs down the road.
int f0(int x) { return x; }
int f1(int x, int y) { return y; }
into a single function, so at runtime, (void(*)())f0 and (void(*)())f1 would compare equal. AFAIK, you are not guaranteed by the standard that functions of different signatures would, when their addresses are taken, result in different pointers, so it's not technically a bug... but it's quite surprising, and ruins certain tricks.