Broken the patch submission process again somehow

[atomrab]

I noticed that there was a period in the DINAA authority that didn't have integer values because it had been submitted as BP dates with no ISO equivalent (starting in "2500", ending in "300"). I made a local IDB, imported the DINAA dataset, and corrected the issue.

For some reason, when I submitted the first patch, only one of my changes had been saved (to the end date). I approved the merge before I realized this, with this result:

When I saw that, I made sure that the dates had been added to the start, removed the "BP" from the end (it wasn't the original statement, but an inference), made sure the explanation for the change in the editorial note was present, and submitted a new patch from the same IDB (which was probably my mistake); it removed the BP but didn't add the start or the editorial note, and I rejected it:

Then I went back to the same collection, made sure I'd saved everything properly so that it looks like the following, resubmitted the patch, and got the same result -- so I rejected it again.

At this point I started to assume that the problem was the old mistake of making a local collection and submitting multiple versions (although here the URIs were already in place, so I assumed it wouldn't matter). So I made ANOTHER backend, imported DINAA from the Canonical dataset, and edited this period there.

When I tried to submit the patch from this new backend -- which should have a difference in both the ISO start date and the editorial note -- to the Canonical dataset, no changes were detected:

But in the History for that datasource, there's a record of the patch and those actual changes:

Here's the patch (I made a third, clean, new datasource for this, to make sure I hadn't accidentally imported changes from the earlier local one rather than the canonical). Any idea what's going on?

dinaa_patch.txt