On narrow unicode builds:
unicodedata.category(chr(0x10000)) == 'Lo'  # correct
Py_UNICODE_ISPRINTABLE(0x10000)    == 1     # correct 
str.isprintable(chr(0x10000))      == False # inconsistent

On narrow unicode builds, large code points are stored with a surrogate pair.  But str.isprintable() simply loops over the Py_UNICODE array, and test the surrogates separately.

There should be a way to walk a unicode string in C, character by character, and the str methods (str.is*, str.to*) should use it.