Yeah I don't really get this either. The author suggests asking a lot of detailed questions about which arm participants thought they were in. What happens if participants (as you would expect) make very accurate guesses?