On linux I have a path which python reads as...

/data/test/num\udce9ro_bad/untitled.blend

os.listdir("/data/test/") returns this ['num\udce9ro_bad']

But the same path cant be given to the C api's Py_CompileString

Where fn is '/data/test/num\udce9ro_bad/untitled.blend/test.py'
 Py_CompileString(buf, fn, Py_file_input);

...gives this error.
UnicodeDecodeError: 'utf8' codec can't decode bytes in position 14-16: invalid data

From this pep, non decode-able paths should use surrogateescape's
http://www.python.org/dev/peps/pep-0383/