palomar: prod errors burndown (#481)
I took a peek at prod error logs and found some issues:
- deleting search documents is failing, because the "document ID"
(docID) was formatted wrong, because the `rkey` parameter is actually a
full path, not just the record key part
- many actually-valid datetimes were not getting parsed correctly,
resulting in posts not having correct createdAt, likely resulting in
them not appearing with expected ranking in search results
- many HTTP 404 errors on fetches of base path; not really a big deal
but might as well return healthcheck
I haven't tested these much yet, we should probably deploy to staging
(which consumes the prod firehose, IIRC) and observe that for a bit.