I suspect the Histograms that you looked at are for the entire DD reports, not just the Thon profiles,
Correct, the histograms only count full weeks. The Thon report is a subset of the weekly report, so anything listed in the Thon report is already in the weekly.
They account for the number of errors corrected from last week. So if there were 5000 new errors, and 4000 were fixed from last week's report, the histogram will show a delta of +1000. Negative values are also possible.
which is why you are seeing profiles that have not been edited.
Not quite. The histograms don't show individual profiles, so I can't be seeing it there. I'm seeing profiles which have not been edited during the week in the weekly reports. See here, this week's 586 New report. Looking at a random selection of these profiles:
I suspect that Ales is mostly scraping FindAGrave profiles for changes on an as-needed basis, both to conserve processing power and to prevent hammering FindAGrave with traffic. (After all, most memorials don't change super frequently.) It appears these merges happened some time before the Thon and are just now popping on the report.
What about the Thon caused these to pop up this week is unclear. Perhaps someone edited a nuclear relative, which triggered a refresh of the FindAGrave data? However, we can discern that the above folks are not on the report because of being edited or added during the Thon.
The same holds true for the 857 New report. Note that there's a huge spike in this error this week, the first time since stats have been recorded. Again these don't look look like edits made during the week of the Thon. Some external factors with FG must be at play.
I hear you on seeing an increased number of new FG errors. But is that unusual when considered beside the large increase in profile creation and increased site activity?
The FG error rate for new profiles has varied between 8.98% and 17.76% in 2024, with a mean of 13.53%. For the two weeks of the Thon, the figures are 15.20% and 13.87%, or "about average." The math is available on this 2024 FindAGrave Error Metrics GSheet.
tl;dr: It's likely any other event that increased site activity to a similar amount would have generated a similar number of FG errors.
[Edit: line breaks for readability]