I want to love it, and I don't take open source projects like this for granted. But during my last production upgrade I chose to decommission Litestream in favor of a dumber, less granular solution using sqlite3_rsync and nightly backups because there is no point in using a backup system that is not rock solid.
By default. Generally your primary database is in a completely different failure category than a kubernetes node running an ephemeral workflow pod.
synchronous_commit = on PRAGMA synchronous = fullThat's distributed workflows :)