You’re right, that helps. Obviously there must be a bug in the way the displacement map is read though. Probably a rounding error from not sampling in the middle of a pixel. Working with image samplers a lot in coding as well, I know it can happen quite easily that you actually sample a pixel on the edge and then there might be rounding errors. You actually have to offset by half a pixel to make sure you sample a pixel in the center of it.
I hope someone from McNeel can take a look. At those scales being off by just a pixel can mean a meter to dozens of meters difference.
I noticed that setting repeat at 0.999 still works, but at 0.9999 the error still occurs.
