I have been an AP exam reader (scoring free-response questions) for just over a decade now and we have moved from all of us reading on-site to about 40% of us in my subject reading on-site and the rest reading from home.
We all score responses that have been scanned from paper booklets and we have benchmark papers provided for reference, so theoretically we have identical levels of training, but what the at-home readers do not get is the regular interaction with other readers and being able to consult with a table leader in person immediately if there is a response that is tricky to score. Teamwork and multiple sources of input are invaluable and I believe it makes us more accurate.
If an at-home reader wants feedback, a response is placed in a temporary hold and then hope the table leader gets back to you in a reasonable length of time (sometimes the at-home readers and their assigned table leaders are 2-3 time zones apart).
Validity papers are slipped into our queue on occasion to ensure that we are staying on rubric and scoring responses as they should be scored, and the at-home readers for the question I scored this year were doing significantly worse with those than the on-site readers, so much so that those of us on-site had two separate brief meetings with a member of the rubric team about one part of the question, and we all looked at each other like "we aren't the ones screwing this up...our validity paper scores are excellent." They were talking to the only people they could, but it was misdirected frustration.
Our at-home readers were also performing at about 1/3 of what they were expected to do in terms of getting responses scored, which was an immense source of annoyance for those of us on-site. The screening process for who reads from home must be improved. I am sure there were at-home readers who were not yet finished with their school year (we scored papers during the first week of June) and waited until the weekend (the last two days of scoring) before doing any serious work. That is not acceptable.
All of that to say that I think that, on the whole, the quality of scoring of our questions has suffered because of the change to an at-home/on-site hybrid model, but it is not going away. I know there are readers who do a tremendous job from home, but they are in a distinct minority. If College Board had their druthers, we would all score remotely. I will never score from home again, so if it goes 100% remote, I'm done. Apologies for the semi-vent...I'll just put this particular task as one that, in my opinion, has not seen improvement with WFH introduced.