Reverse sync-to caused repos to desynchronise(?)

Here’s a scenario:
1 local filesystem repo (“onsite”) where snapshots get created,
1 remote sftp repo (“offsite”) synced from onsite using sync-to.

It just so happened recently that the onsite volume got corrupted.
So, created a new volume (same FS, etc.), used sync-to to get the offsite copy back to onsite. Worked fine AFAICT.

Created a new snapshot in onsite repo.
Trying to sync it to the offsite again (exactly the same cmd as always before) and this happens:

  Found 26311 BLOBs in the destination repository (609.1 GB)
  Found 32494 BLOBs (754.7 GB) in the source repository, 32494 (754.7 GB) to copy
  Found 88 BLOBs to delete (280.9 MB), 0 in sync (0 B)

Then it proceeded to push the entire onsite repo again to the offsite one (which I interrupted for now).

What gives? The repos only differ by 1 snapshot, the rest should (and appears to, checking snapshot list, etc.) be exactly the same.