Only part of a lay-up is glass. The rest is resin, which is a plastic, and may have a higher rate of expansion then glass. An earlier post by Terry indicated a rate of .030"/ft. I don't know if that is correct, but let's assume it is. I'm not sure what the length is in a JMP T-33 between the servo and elevator horn, but I believe the servo is located in the front of the model, so it could be 3 feet. If Terry's #'s are correct, that would mean a change in length of .090". If the control horn on the elevator is short, that would mean a significant change in the trim. Again, I am not saying this is what was ocurring, just one possibility.

Another very plausible cause mentioned by jason is the outer sleeve of the pushrod. If it is the typical plastic sleeve, it will expand a large amount with heat and will cause a change in length of the pushrod. My bet is that this is the cause.

IMO it's a jump to think that because a gyro may have a trim change with heat that a servo will also. Servos do not have the piezo-electric crystal that is in a gyro. It's my understanding that it is this crystal that can cause the trim drift.

Aptar, just take a heat gun to the servo and see if the center changes. Should be an easy test to conduct, and may eliminate that as a cause. I still feel that the real solution is to reduce the heat build up. One thing to look at is if there is enough source of intake air into the fuselage to cool the tailpipe. Another would be the back of the pipe. Is there an augmentor nozzle at the back to induce flow over the inner tailpipe? The end of the inner pipe should be some distance in front of the end of the outer.