sarge wrote:Wayne, I suspect the problem is an extreme example of the classic reverse-curve problem.
Well, I assembled this reverse curve with a three-rail O-27 switch just to test the classic reverse-curve problem. So far this 2-10-0 is the only piece of equipment that has a problem. Actually, the locomotive, while the cause of the problem, has no problem going back and forth through the S-curve without the tender attached.
sarge wrote:Try moving the switch to the other end of the straight section so the tender can go back normal before making the reverse turn.
I am sure moving the switch to avoid an S-curve would solve the derailment problem. The S-curve is there to test the S-curve problem.
sarge wrote:You have one rigid wheelbase defined between the bolster pins of the tender and a second defined between the first and fourth driver set, and you are trying to pull the first with the second (by a whole bunch of flobbery bits in between) that are at a very sharp angle from the axes of travel defined by each.
Correct, the angle required by the drawbar and tender connection is the problem.
"Flobbery" is not a standard English word. Still, it could be a non-standard or colloquial form of the verb "flobber," which means to sag or wobble, or an adjective form of that verb describing something that exhibits this characteristic. The word "flobber" itself is obsolete, recorded only in the Middle English period.
The drawbar is held in position with a bolt. There is a spring under the drawbar. The spring keeps the drawbar parallel with the bottom of the engine. The drawbar has curbs that prevent its swing from exceeding 90° (±45°). No flobber with this setup.

The tender has its drawbar attachment riveted to the truck. If it could swing independently of the truck, I am sure there would be no problem going through the O-27 S-curve. The engine specifications say, minimum curve: O-31. Since the engine-tender almost makes it through an O-27 S-curve, I am pretty sure there would be no problem going through an O-31 S-curve.

sarge wrote:This is why it is considered good practice in track design, prototype, scale model, and toy, to lay in a straight between the curves in a reverse curve or crossover equal to the length of the longest piece of stock intended to traverse the form.
I know, but what's the fun in that?
One of many possible S-curve problems, limited drawbar swing.
