Data Dump keys sometimes have wrong format

[cdrini]

The format of some of the keys is wrong though. E.g. instead of /authors/OL123A, it sometimes lists /a/OL123A
This looks like it was a regression between the 2021-10-13 data dump and 2021-11-30 data dump

Evidence / Screenshot (if possible)

$ curl -L 'https://archive.org/download/ol_dump_2022-03-02/ol_dump_authors_2022-03-02.txt.gz' | zcat | grep -F 'OL3965446A' | head -n1
/type/author    /authors/OL3965446A     1       2008-04-30T20:50:18.033121      {"name": "I. Meyerson", "last_modified": {"type": "/type/datetime", "value": "2008-04-30 20:50:18.033121"}, "key": "/a/OL3965446A", "type": {"key": "/type/author"}, "id": 16613668, "revision": 1}

$ curl -L 'https://archive.org/download/ol_dump_2022-03-24/ol_dump_authors_2022-03-24.txt.gz' | zcat | grep -F 'OL3965446A' | head -n1
/type/author    /authors/OL3965446A     1       2008-04-30T20:50:18.033121      {"name": "I. Meyerson", "last_modified": {"type": "/type/datetime", "value": "2008-04-30 20:50:18.033121"}, "key": "/a/OL3965446A", "type": {"key": "/type/author"}, "id": 16613668, "revision": 1}

$ curl -L 'https://archive.org/download/ol_dump_2021-11-30/ol_dump_authors_2021-11-30.txt.gz' | zcat | grep -F 'OL3965446A' | head -n1
/type/author    /authors/OL3965446A     1       2008-04-30T20:50:18.033121      {"name": "I. Meyerson", "last_modified": {"type": "/type/datetime", "value": "2008-04-30 20:50:18.033121"}, "key": "/a/OL3965446A", "type": {"key": "/type/author"}, "id": 16613668, "revision": 1}

$ curl -L 'https://archive.org/download/ol_dump_2021-10-13/ol_dump_authors_2021-10-13.txt.gz' | zcat | grep -F 'OL3965446A' | head -n1
/type/author    /authors/OL3965446A     1       2008-04-30T20:50:18.033121      {"name": "I. Meyerson", "last_modified": {"type": "/type/datetime", "value": "2008-04-30T20:50:18.033121"}, "key": "/authors/OL3965446A", "type": {"key": "/type/author"}, "revision": 1}

Proposal & Constraints

Related files

Stakeholders

[cdrini]

P1 because blocking #6119 and because it likely breaks anyone's workflow who's using the data dump.