one of my notebooks is corrupted

I cannot open one of my notebooks. The message states:

Unreadable Notebook: /path/lmo_processing.ipynb UnicodeDecodeError('utf-8', b'\x92 \xf9\xa9\xc9\x92\t\xe7<\x1b\x00\xd9\x9d\xba\xb3\xd9\x92 \xf9\xa9\xc9\x92\t\xe7<\x1b\x00\xd9\x9d\xba\xb3\xd9\x92 \xf9\xa9\xc9\x92\t\xe7<\x1b\x00\xd9\x9d\xba\xb3\xd9\x92
(several thousands of lines like these)
\xf9\xa9\xc9\x92\t\xe7<\x1b\x00\xd9\x9d\xba\xb3\xd9\x92 \xf9\xa9\xc9\x92\t\xe7<\x1b\x00\xd9\x9d\xba\xb3\xd9\x92 \xf9\xa9\xc9', 0, 1, 'invalid start byte')

The terminal message says:
[W 12:20:14.952 NotebookApp] 400 GET /api/contents/logs/lmo_processing.ipynb?type=notebook&_=1516015213808 (127.0.0.1):

I made a copy of the notebook and tried to load it in vi, etc. Result:
� ��ɒ �<^[^@ٝ��ْ ��ɒ �<^[^@ٝ��ْ ��ɒ �<^[^@ٝ��ْ ��ɒ
Even my system reports it as byte not text file. It is the only faulty notebook out of hundreds. Can it be that some faulty backup operation make a disk error, or is it something you have seen before?

Begin with the reported GET /api/contents/logs/lmo_processing.ipynb request and the UnicodeDecodeError shown when loading the notebook. The issue names no source file, test, version, or reproducible setup, so a confirmed reproduction and an agreed handling of the unreadable notebook are needed before implementation can be scoped.

Written by the indexing model from the issue text.