After determining the source of my derailment was a sudden twist to the track, I spent a few days pondering how to address it. Should I shim or should I sand? Which rail needed to move? The Trackwork Torsion Testing Tool (TTTT) told me where the problem was, but not exactly what the problem was. It was time for another tool!
The Transverse Trackwork Trim Tool (oh dear, four Ts again!) is a simple cradle for a level that holds the level clear of surrounding trackwork and some scenery so that I can identify if one rail is higher than the other. It was a matter of half an hour of design work and five hours of printing overnight to get one, and it immediately proved its worth.
Using the new tool I quickly determined that when I sanded the ties in this curve, I had super-elevation in mind, and made the outer rail higher than the inner rail. At the north end, I appear to have made the transition smooth enough, but at the south end, it is more abrupt. Sadly this meant sanding rather than shimming.
So, I accepted the fact that there would be a major service disruption three weeks before the regional NMRA meet layout open house. Because the track is behind the roundhouse and turntable, and because I need to repair the turntable as well, I removed that also. The old soldering iron quickly loosened the Pliobond holding the rail in place, and I was able to lift the rail out of the way. When I did so, I was alarmed to discover the feeder was severed; it was probably for the best as it gave me more room to sand.
I then made a little sanding block and carefully removed a surprising amount of material in the transition zone. I discovered that I could hold the rail in place with a track gauge or two, and test it with the TTTT (TM) until the needles stayed much closer together, indicating a more gentle twist.
When I glued the rail back in place and hooked up the new feeder, I was delighted to find that 1120 stayed on the track!
Pembroke:87 
Glad to see that you remedied this, albeit with some effort. Cross level has been an issue that’s bedevilled the prototype almost since the inception of railways. On loco’s with some sort of suspension system, the suspension adjusted how the loco sat on and traversed the rails. As modellers, we’ve a challenge working within the tiny spaces on a model to optimise a loco’s traverse down the track. Your Trim Tool is almost the opposite of the common track level in design, but almost identical in function!
https://pwayblog.com/2016/01/25/how-is-the-cant-measured/
Steve Lucas
Thanks for the link. It might be interesting to add a screw to the Trim Tool to enable a known cant to be found.