Good news - following a reset of the receive-side database via the /rest/system/reset endpoint, upon unpausing the folder it quickly reported “Up to date” instead of “Local additions”.
Better still, upon setting up a similar send/receive folder pair, that one worked first time.
I still wonder whether this was due to only enabling ‘ignore permissions’ after the initial sync calculations, but that would fail to explain why the problem only applied to about two-thirds of my files, when all seem to have the same permissions on the sending side. It’s interesting that @gyver also seems to have had around a 60% mismatch rate - could there be some value getting abbreviated or quantized in some way that ends up being wrong 30%-40% of the time? (Maybe only with v1…)
The mogration is in so far relevant as we trust the old database. So we might have simply copy pasted the bad state into the new schema. The real bug is that only a reset of the folder can fix this.