The first might be "Let's replace all use of tp_* with hash lookups", which I suppose would be closed quickly, amidst a storm of muffled giggles. But at least by then you'd know why the current implementation works the way it does.

The next might be "__getattribute__ mechanism has surprising behaviour in common use-case" (which itself would imply having a use case where any of this mattered). This one might actually result in some productive discussion regarding a fix.

Your second reply is basically a doc bug, one of "__getattribute__ optimization is not specified in the language reference" or "__getattribute__ optimization causes non-comformance with the language specification", either way, you would get the attention of the few people who can actually help you, rather than attention from the many more who can't