I think it might just have been that the capability makes the pattern matching unbounded in human complexity. The right side's expression is already practically

I think it might just have been that the capability makes the pattern matching unbounded in human complexity. The right side's expression is already practically unbounded in complexity, making the left side too flexible becomes insanely difficult to figure out. And when the Haskell community is telling you that some feature is too difficult for humans to use sensibly it's best to listen. I think simple pattern matching is useful, especially type-based pattern matching, including which element of a sum type is matched. But you want to watch out for how complicated it gets on the matching side. Complex matching code is a very strong form of coupling. It may be a good thing but I think as a language designer you don't want to overpower it because of the coupling it introduces as the power gets used.