The constraint makes it much more challenging to use. For example, in a generic method that supports working with arbitrary Ts and in which you want to accelerate the handling of some, you can't access Vector*<T>.IsSupported, you can't write routines that first check whether T is a known / supported type and then use VectorX<T>.Create(T value), etc.

According to @tannergooding, this constraint used to help improve codegen but with improvements already made to the JIT, we should be able to remove it without negative impact on code quality.