Finally reviewed Stefan's latest patch. It mostly looks great to me. Aside from a few minor cosmetic quibbles, the only substantive change I would like is to expose the Py_buffer len field as "memoryview.size" instead of "memoryview.len" (to reduce confusion with "len(memoryview)" and to encourage a conceptual link with sys.getsizeof()).
As noted in the review, I also think fixing #12384 should be fairly straightforward and it would be nice to have that working when the change is applied.
(Kristjan: I added you to the nosy list, since you commented on another issue that you had been poking around in the memoryview internals and noticed a few dodgy things. Stefan's work on this issue is aimed at comprehensively addressing those problems).