It looks pretty easy: _readdir_pyx in _read_dir uses a statvalue of None for files that disappear(?) between the readdir and the stat, but UTF8DirReader.read_dir doesn't handle it.
We could handle None but I'd actually question whether we're doing anything useful by returning data about these files- why not just skip them and treat them as totally gone?
It looks pretty easy: _readdir_pyx in _read_dir uses a statvalue of None for files that disappear(?) between the readdir and the stat, but UTF8DirReader. read_dir doesn't handle it.
We could handle None but I'd actually question whether we're doing anything useful by returning data about these files- why not just skip them and treat them as totally gone?