This is exactly the situation Stefan was talking about: "However, I still think that Serhiy is right: this change would break code, and in particular test code that compares XML output. Having to deal with two different "correct" serialisations in tests depending on the Python version is annoying at best."

He says comparison should be done with true canonicalization (C14N).  Does that exist in the standard library?  With sorted keys, we had a form of good-enough canonicalization.  Now we don't.