Update on two things:
I was wrong about flange depth. The driver flanges on this loco are ~2.2 mm deep, vs my Lionel postwar Berkshire, ~3.2 mm. Thinking about baseline situation, with rear 2 driver axles carrying weight over a fairly modest “hump” in the track, it wouldn’t be hard for the 2.2 mm flange on the lead drivers to be lifted clear of the railhead so it could come down on the wrong side of the rail after that hump in the track. Did not directly see that happen, but it may account for the problem.
Pilot truck derailments continued often enough to be a nuisance. I noticed there seems to be significant side to side slop allowed in the rear (pivot) end of the truck. That could allow the axle to run at a larger angle to the rails when it is caused to do so for whatever reason, so it might find a flangeway or frog to pick. I tried making a spacer out of a 1/4" nut drilled to 7/32", filed to fit on each side, plus on the rear-end side so that both screws lined up when done. Now there is miniscule side movement. A number of circuits at track speed of 45 mph this morning with no derailments. That slop may be needed for some tighter curves. My minimum radius is about O62 (31" r), and it doesn't seem needed for me. Photo of the fix below. The difference with Lionel Berk doesn't look huge, and most folks obviously aren't having a problem, but this seems to be my solution.
