I think this is a new problem, because with the new track textures I added quite a bit of high-frequency data to the aplha/reflectance channels in both the detail maps and the base texture. If I had to guess, all that fine detail is getting biased towards white when the Intel DDS exporter is generating mipmaps. The Intel exporter doesn't give any tuning options to prevent this. NVidia Texture Tools does, but that tool was the one that was exporting weird DDS formats that were (possibly) causing crashes under Win7. If I can't get a well-formed DXT5 file out of the NVidia tool, the workaround is probably hand-adjusting mip-map alpha channels (ugh).
Interestingly, I don't see this on my end, but it may be because I have anisotropic filtering cranked way up in graphics settings, which could be smoothing the mipmap transitions out more than in your screenshot.
ETA: I was able to replicate the track "shadow" by reducing anisotropic filtering quality. You can see the mip-map transition is still slightly evident at 16x under close inspection, though:
View attachment 398633
Hopefully I can tune this out by adjusting the mipmaps.
To further elaborate on the Merge Meshes x CSP windmill animations issues, both 3DSimEd round-tripping and car-format ks5 files from ksEditor prevented the crashes, but the car ks5s had some strange coordinate-space differences from a track ks5 that alternately caused the turbine blades to reorient to the global Z axis instead of their internal Z (leading to weird wobbling rotations that I couldn't tune out) or turning them into flying death shurikens, depending on the CSP configurations used.
View attachment 398631
Are there particular issues with using files from 3DSimEd, or is it just considered low-brow?