Your last line refutes your refutation! If Haskell really can "optimize much more," then it's harder to know from code inspection what the program will actually

Your last line refutes your refutation! If Haskell really can "optimize much more," then it's harder to know from code inspection what the program will actually do at runtime. And if you don't know why your program is fast, then you can't know how to keep it fast. Does this seemingly innocuous change defeat an optimization? Hard to say.