Hardly a "return." Sigil has never really stopped ensuring that the opf is in state that Sigil can properly parse. Nearly every aspect of Sigil's automated processes are predicated upon the opf being in a state that's both valid (RE spec) and useable by Sigil's parsing routines.
The OPF is one of the few places where Sigil's wishes may still trump the user's. Just hopefully only in minor esoterica like this example. I've not run into too many ebook creators who have an interest in preserving technically correct, but functionally purposeless namespace prefixing schemes. *shrug*
I know you were probably posting in a primarily ironic manner. but I wanted to be sure people understand that Sigil's move towards a "hands off" approach does not extend to every single aspect of the OPF.
With regard to the uncommon prefixes in this test sample, I believe I've seen those ns0, ns1 prefixes introduced by lxml parsing where no attempt to properly manage the namespace prefixes is made. Probably some sort of in-house automated script that no one has ever revisited since epubcheck doesn't bark about it.