One of the understated parts of the article was where they compared how the poorly scoring set did on the EXACT SAME QUESTIONS as the previous, better scoring s

One of the understated parts of the article was where they compared how the poorly scoring set did on the EXACT SAME QUESTIONS as the previous, better scoring set was. All pretense and human drama aside, that sounds like a pretty reasonable foundation, and it rates against the poorly scoring students. I don't want to ignore the likely real emotional impact of the technical snafu, but considering the extreme stress the students are already under, I doubt it would explain the degree of difference. Getting actual reliable metrics one way or the other would be difficult, but essential, I think.