If you think back to the first survey (mid April 2020), what did you think was the probability that:
In session 3, the rating scale was wrong (from strongly disagree to strongly agree), and it was cnaaged for session 4, thus the data are incompatible