Are we sure this will always skip the latest one? We could consider explicitly sorting the collection or (IMO better) finding max(created_at) and skipping

Are we going to replace this with the batch_write helper? (D8895)

IIRC @varun had some issues with writing reserved usernames in a loop, but these were larger quantities than this

From query documentation: "Query results are always sorted by the sort key value"

...although later in the same paragraph they say "To reverse the order, set the ScanIndexForward parameter to false.", which wasn't correct in my experience (in both cases the order was the same). I will implement your second suggestion.

We need the returned values to cleanup user_data (it's only in the full backup table and not in index). Barring any errors, there should always be only one backup (the previous one) that is being deleted anyway.