Something not mentioned is what might be happening to the clutch alignment under load. Given the tremendous torque that this motor makes, some twisting movement between the primary and secondary under load could cause a belt to run up against the edge of a sheave and then cause the belt to walk off the primary or the secondary. It would be almost impossible to check the degree to which clutch alignment is affected under full load, but it emphasizes the importance of rigidity in a chassis, especially as it pertains to the rigidity between the primary and the secondary. How still do the motor mounts hold the motor? How still does the jackshaft remain under load? As the primary and secondary are independent of one another, the rigidity of the chassis is crucial to maintaining clutch alignment. You can't easily observe what's happening to the clutch alignment in actual conditions (when there's a real-world load at the track). I'm not saying anyone here is wrong ... just that clutch design might not be the only cause.