This seems to be happening due to floating point rounding errors. These are large enough to be noticeable because you have an insane scene setup: your armature is scaled to 0.2, and then you scale the whole model by 360 in the QC file. You also have the origins of each of your meshes miles away from the geometry.

Once I brought everything within normal tolerances the problem went away.

Someone else had the exact same problem: https://psteamcommunity.yuanyoumao.com/groups/BlenderSourceTools/discussions/1/3277925755450662348/

Detailed models don't play nicely with SMD/VTA import and there's not much anyone can do with it.

Please try again in the latest version of Blender. If it still doesn't work, upload your blend file and I'll have a look.

The bone has an invalid transform. You'll need to replace the matrix (with the python console) or delete and recreate it. Unless you know what you're doing, just delete it.

I'm surprised. There must be a shader doing per-vertex operations seperately from the main render pass? In any case this needs fixing. I'm working on it but it's something of a black box!

The important thing is that both models have the same number of indices and the same draw count. This means that they render equally quickly. The difference in the number of verts means only that more memory is being used.

As for why the vert count is different, it looks like the Blender tools are doing something that prevents StudioMDL from optimising the size of the vertex buffer. This is presumably something to do with normals, UV co-ordinates, or weights.

It's one of the options in the bottom left of the import window.

Looks like the LOD you're trying to import references a non-existent bone. Try changing the Bone Append Mode when you import to Validate Against Target.

Nothing to do with the Source Tools and you can ignore it anyway.

Thanks, I've fixed this in 2.10.2.