I don't like this assert either: this function is part of the public
API, and should not crash the interpreter just because of a trailing \.

To test easily:

import ctypes
decode = ctypes.pythonapi.PyUnicodeUCS2_DecodeUnicodeEscape
decode.restype = ctypes.py_object
decode(b'\\1', 1, None)

This should gently raise a UnicodeDecodeError IMO.