Because a lot of languages can't easily express the semantics of "you're not allowed to call the function this way". I mean, suppose you have a function that takes two integers, but the second must be greater than the first. Most type systems don't have any way to represent that at all, let alone conveniently.

Related: don't write if(arg == null) throw ArgumentNullError; at the top of every function. If it's not supposed to happen then just let it throw the error naturally when you dereference it. In C it's even worse because you turned an easily caught segfault into a silent nop.

If the criteria for it not happening is too complicated to expose to the caller