This slide shows a riddle we solved some time ago with the help of this logfile analysis. Our two machines are identical, but interestingly, we had much less hold-offs and therefore faster treatments of the same field on one machine. Here the logfile analysis helped us track down the problem to different MLC controller software versions (Joerg Bohsung maybe will remember this). Clinac Blue had the older version (4.8), which handled closed leaf pairs differently than the newer version, it did not assert hold offs for closed leaf pairs. We had this suspicion because of the higher error fluence in the periphery of the field. This also explained the faster treatments.