#1097 added crawler errors to the index, and while the PR is working as expected, we should now delete (or hide) old errors to avoid cluttering the web interface.

Here are a few options:

• Add a new janitor command to manually purge the error documents.
• Add a background thread to the API to purge old (e.g. >1month old) docs periodically.
• Add a new web client API to fetch every errors, and only return the recent one (e.g. <1month old) by default, with a more button at the end.

What do you think?