I tried playing with the gltf (at this stage I don't really want to deepdive into Blender and its exporter), and I needed to make the following changes by a text editor:
1. All animations have to get their proper MSTS animation name. In the .s file one needed to name the nodes. Here one needs to name the animations. What you e.g. named as "WHEELS4Action", I needed to rename to "WHEELS4", "ROD19Action" renamed to "WHEELS", "DOOR_AAction" renamed to "DOOR_A", "ORTSBELLAction" renamed to "ORTSBELL", etc.
2. All nodes that you want to be animated by Openrails, at this stage need to be tagged by the "extras" I wrote here previously, e.g. one of the nodes look like this after the edit:
{
"mesh" : 10,
"extras": { "OPENRAILS_animation_name": "WHEELS21" },
"name" : "WHEELS21",
"translation" : [
0,
0.020983219146728516,
-1.0878195762634277
]
},
I set up the code the way that naming the node is not enough (not even needed btw), mainly because at the beginning all the MSTS modellers will just name the nodes and expect they will move just like the .s file moved. But it is not the case, because nodes that are targeted by internal animations should not be tagged for Openrails-managed animation. In your case e.g. the node named as "WHEELS1" should not be tagged with "OPENRAILS_animation_NAME", because that is alredy animated in a different way, with a real animation clip.
Even with these modifications the majority of the rods are not moving correctly. I think it is because you did not include animations for all of them. Please note, it is not enough setting up constraints for them in Blender, because there are no means in the gltf to describe these constraints. You need to include keyframe animations to all of the rods. Did this work correctly by exporting to an .s file? I haven't actually changed how the animations are processed. You can debug the exported animations by the Windows built-in 3D Viewer. However you might want to merge all the rods-related animations in to a single one. There are currently 45 animations defined, as also shown in the 3D Viewer, which is hard to debug. The 3D Viewer would play for you the whole wheel-rods animation together, if you merged them into a single one.
I also tagged the bogies, and they turned into the right direction in Openrails as far as I noticed. The geometry is not exactly right in small-radius curves, like the one when starting at "columbia falls set out (traffic).pat", but this locomotive should not run these small curves anyway. I'm not sure if changes are worth in Openrails to align this more perfectly. Changing the relevant code might cause more problems than it solve, given that we do not have a test suit of various cars collected, to check if a change corrects problems everywhere, or while it fixes something, brokes something else at the same time...
Ah, and I forgot to mention some important things:
1. glTf is oriented in a way that +Z is forward, +X is left. This is different to the .s orientation.
2. The base color texture should be in sRGB color space. The .s file was linear RGB.