To put some scale on the amount of data here.
Given a fairly typical assumption of an average from Yaleof 20 hours per linear foot this one regional collection would take 17,000-34,000 person hours or about 16.5 (2080 hour) person years to run through.
Obviously it has an index, and if your search uses that index it will be more efficient and other tools could reduce this time even more.
While this is probably the largest regional collection this is not all of the data that would have to be run through if you are looking for terms that aren’t the index. The time complexity would not be linear to correlate these either. Nor would it “prove” outside of official requests where this happened.
Sure targeted studies can show that it was uncommon or even rare, but I have already acknowledged that. And no, changes weren’t made out of spite etc…
But yes, if you can’t point to a study that did more than look at the manifests we will be an an impasse. Simply no careful research has been done to prove an absolute claim that it NEVER happened even through clerical error.
Feel free to provide a cite that documents even a claimed detection power of the applied methods or heck even offers a confidence metric and I can consider the null result as evidence of absence.
I have tried to find any such study myself but have failed, but confirmation bias will do that to you and I would appreciate it if I someone can share a link.