The only firsthand knowledge I have with this phenomenon is RGS #20. She rolled over in the early '40s and must have hit something very hard on her left side just in front of the dome. The second course was flattened by about 1 1/2". This far exceeded the out-of-round tolerance allowed for a new boiler of plus or minus 1%. Repairing this kind of damage is a big job and the RGS obviously decided not to and put her back in service until they quit. When we hydro tested her prior to starting repairs, we put dial indicators around the boiler in the area of the flat spot. The boiler flexed a bit less than 1/16" for every increase of 50 psi. This flex occurred within about 12" of a neutral point which amounts to considerable movement or bending of the steel at the neutral point. How many times can you bend a coat hanger before it breaks is what this amounts to. I have often wondered at the conversations that might have occurred in the shop when a wrecked locomotive showed up for repairs. Taken in the context of today's anal-retentive verbage in regards to code and repair irregularities, I am sure those discussions would be an eye opener.
Our GW #90 rolled over on its left side in Colorado and it still carries the dents in the shell where the air compressor mounting studs absorbed the brunt of the force. Her shell is still round within tolerance - unlike #20.
I have been amazed at how locomotive boilers survive off-track excursions. I would love to see RGS #455's boiler or for that matter any C&S locomotive. My impression is that many, if not all DSP&P/C&S locomotives spent at least part of their careers rolling down hillsides. Obviously the smaller the boiler, the less tendency there will be for damage to the boiler proper simply because of size.
I would agree with the other posts in this thread regarding piping and connections which usually do not fair very well.