You mentioned data logs telling the story. That's the real change from guessing. What was the specific data point that finally showed you the problem? Was it a clear spike in line pressure during the shift, or something more subtle like a slow solenoid response time that you could only see on the graph?