The BEEP saga continues.....
I wanted to swap shells on one of my chassis to run a Chessie shell I have. Since this topic popped up again the other day, I spent a little time checking rollers and clearances on 027 tubular.
IN MY EXPERT OPINION.........
THE ROLLERS ARE TOO LONG AND HIT BOTH THE CENTER RAIL AND INNER POINT RAIL ON TIGHT RADIUS TURNOUTS!
The rollers on the test chassis were clean and free of any scars from sparking before I ran it. (Pic taken after a test run) The scars are on the inner side of the rollers.
I did a test by covering the inner curved point of the turnout with electrical tape. NO SPARKS
Looking at the position of the rollers on a curved 027 track section. Unlike articulating trucks, the rigid wheels and chassis don't allow the rollers to remain centered on the center rail on a curve.
And when on a 5122 switch. The gearbox has adequate clearance. The roller in the background is touching both rails at the apex of the curve.
I took the rollers off the chassis and ground down the ends to give them some taper. Sparking was reduced but still not eliminated.
I then went to the source of the problem....the switch. I've used clear acrylic polyurethane in the past as a rail insulator so I applied 2 coats to the inner point of the turnouts. Results are better than before and more acceptable to me. I don't know if the Kapton tape John mentioned would be an option but I'm open to try other solutions.
Overall a productive hour or so on a rainy Saturday afternoon.
Bob
PS- Looks like I will need to add some shielding to the inside of the shell to block the light bleed. 
