The current hack isn't necessary. I didn't review eryksun's patch
in depth, but on the surface it looks good.

I don't think the other issues you mentioned are closely related:
The cause here is that the obj fields of both the second memoryview
and the second mbuf are manually set to the address of the first
memoryview, which plays havoc with the rather complex deallocation
scheme.

I've opened #25524 in order to make it easier to create views
with detailed request flags.



As a practical matter, I think we should just revert the ctypes
change for 3.4 and 3.5 (after verifying a second time that
PyObject_AsWriteBuffer() is exactly the same as in 3.4.0).


Then we use the new function from #25524 for 3.6, which should
alleviate concerns about writablility (not all extension writers
read the docs:) etc.