Wow, this seems like the missing piece for making SQLite a viable option for using it as a production database for low scale environments.

Given that it's using S3 for storage which AFAIK doesn't have compare-and-swap operations, I wonder how it deals with multiple processes trying to backup to the same directory. Is there a risk of two processes clobbering data?

I've had luck with online backups using Roger Binns' APSW which exposes the same functionality. https://rogerbinns.github.io/apsw/backup.html