Did you save purge output or run list-broken-files or notice what versions the output mentioned? Recovering by purging files gives an example (with more context) of what you might have seen.
It would be interesting if versions 1-6 needed file purge. That’d be consistent with previous idea, however previous idea also thought 1 source file had gone bad, and possibly purge found three.
That was a bit more work than I had hoped. I’m still looking for your server logs from March 14th which was maybe the start of the problem. If pastebin continues to not work, you can paste here.
The March 14 Event Viewer VSS error and MicrosoftEdgeUpdate.exe crash then are interesting. Possibly there was some general system problem? That’s beyond me, but I can look at Duplicati.
I’m guessing 13th (job log) was OK, but you can look at the Complete log (if you like) for oddities.
There is a small chance that further analysis (especially if a developer expert helps) can make a little more out of this situation, but info is needed such as March 14th logs, purge notes, maybe a database bug report to see how things are after recreate. Comparisons across recreate are hard because everything moves, but might be possible. The purge also throws in some more changes.
It’d probably run a little faster, and with less concern about any lingering issues, but my slow SQL in DB Browser for SQLite (so not even Duplicati) was probably separate. Query after 4 days said:
COUNT(*)
410008
which is what I’d already wrestled out of the database manually earlier, based on a shorter query. Good to know that the full query gets the same answer, and both are still short of 410009 by one.
One approach which might serve your need as well is to keep the questionable backup awhile, if you have space. Starting fresh loses history on everything, and also loses history of the problem.