It’s a binary-encoded merkle tree

It's in a sqlite database. Unison doesn't prescribe how those are passed around, but since everything is content addressed there are never any collisions. You can always merge two of them into one without having to resolve conflicts and without breaking anything.

The way I intend to use it is that users periodically fetch the codebases from all of the developers that they trust. This has no immediate effect because you're still calling the same function that you're used to. So the sync can be very latency tolerant.

Then, when you decided to use the new version of something, you call the new hash. If something goes wrong, you just go back to calling the old hash--atomic upgrade/downgrades, no merge conflicts.