I shouldn't do this, but my inner Don Quixote is egging me on.
I'll preface that in case you didn't know, it is no secret that I think all forms of radial thread pins are the biggest, saddest joke in the cue industry.
A properly designed threaded joint (construction screws aside, but our cue joints share nothing with them) should not rely on a tight minor diameter to stay together. A proper screw joint should act like a cable pulling two objects together--the forces between the components should be what keeps the two pieces together. This is called a two-force member. Screws should not be subject to shear (side) forces.
That said, there are only two reasons that screws loosen. First, if the screw is unloading and the tension is lost. Second, there is not enough friction in the joint (in our case, most of the friction should be between the joint faces, not on the minor diameter of the screw.
In the first scenario, it is possible that faces that aren't flat can cause some unloading of the screw. This would mostly be if one or both faces have any crown and the outer edges are not seating. When hit, the joint will flex and cause weirdness in the pin loading.
In the second scenario, there are a few factors. First, if it takes a lot of torque just to screw the pin into the shaft, you reduce the available torque to seat the joint since we are limited by our grip strength to provide the torque. Second, if you have slippery materials on the joint faces, this will greatly reduce friction. I had this problem recently on a cue, I scuffed the faces back to a 400-grit finish and the problem never resurfaced (pun not intended). Finally, it may be a bit of a stretch, but since you are seeing it on CF, there is a chance that the natural frequency of the CF shaft is high enough to reduce friction in the joint just from the vibrations--this phenomenon is prevalent in aerospace, in flight many friction forces go to nearly zero due to the vibration, and FAA requires all friction dependent fasteners (i.e. screws) to be secured with a secondary method (often wire through the fastener.
Inspection will be your friend. I would suggest that a good magnifying glass or digital microscope (I just got one for $35 from amazon) will be helpful. Inspect the joint faces for any witness marking which may show where the joint is seating and where it isn't.
