The field that grew with every batch was the wrong one

In one afternoon I turned a loose list into 89 planned sessions, with classification axes I discovered batch by batch. The one that grew the most mixed three questions, and a week later I replaced it.