Late answer, but googling led me here, and I figured it may lead others here too.
Your tessellation is moving the final location that the pixels of the terrain land, in world space, up and down. But raytracing doesn’t do tessellation. It just uses the original shapes.
So when we try to cast a ray from one of those shifted pixel locations, if it went down into the ground from where the mesh was, it will immediately hit the mesh, and “self-shadow”. No light can get past the ground mesh to the shifted ground pixel.
There’s a variable, “r.RayTracing.NormalBias” (default 0.1) that some people have used to hack around this. NormalBias is saying “for all raytracing casts, shift upwards along the normal by this amount”. So you can make all of the rays start higher up from where the tessellation pushed the ground down. But it changes the way all rays are cast, so keep an eye out for whether it adds other artifacts elsewhere in the scene.