Earlier, Percy sheared off a crank pin. So, conjecturing that this meant the fast starts and stops were causing the wheels to get out of quarter sufficiently to put some force on the crank pin, I continued to replicate #10 and #622. Both of those have brass crank pins, but I turned some for Percy out of steel.
Now, running the improved Percy back and forth like a kid in a soccer drill failed to break anything. I really tried, but I tired of twisting the knob on the controller before Percy would even hint at failing.
Just about the most violent thing that ever happened to either #622 or #10 was when their crank pins unscrewed and a rod came loose. Typically one wheel would turn the wrong way, and the engine would bind, or the dangling rod would jam on a tie. Either way, derailment was certain.
With a little encouragement – about as much as getting that soccer player to stick their tongue to a frozen sign post – Percy was convinced to drop one end of a connecting rod. It still proved remarkably resilient, and ran back and forth quite a few times before finally jamming on a tie.
Sure enough, when I reinstalled the connecting rod, Percy bound up tighter than a pair of wet swimming trunks. Sighting through the spokes, the driven wheel was a full spoke out of quarter – 30 degrees! This confirms that even the small motor in Percy is capable of breaking the friction fit between driver and axle.
Amazingly, I could return Percy to running condition by sighting through the spokes and re-quartering the slipped driver. When I did, the little engine behaved just as before: the soccer-drill-running still failed to produce sufficient force to make the wheels get out of quarter.
Pembroke:87 
I find this puzzling. I am wondering whether the quartering or the fit of the connecting rods are the real issues on 622. Because your testing is (for me at least) leading to the impression that the tolerances are not critical (and so the care you have taken should be more than ample, and the wheels, pins and rods should not be the cause of problems). If that conclusion holds up, maybe you can devise testing to dissect the relation of the motion and drive rods as well?
I have one more thing to try before working on the main rod. Stay tuned!