This is clearly a bug. First of all, occlusion and translucency sorting in the raycast is not a trivial thing, the renderer is computing different kinds of depth and nebula density gradient information for optimizing ray-casts (distance fields, …), all of this is directly on the gpu - unlike the regular objects that are stored on the cpu (3d mesh / poly or a primitive (that are especially gpu-cpu optimized for intersecting ray-casts, as seen in the PS4 Dreams project linked in the topic)), therefore gpu-cpu interaction is always a tricky thing, but I might be slightly wrong with this explanation. Yet our renderer is not optimized for very detailed occlusion calculation and is meant to be scaled quite large. But anyways, Schlabber will need to have a look, maybe it is just a vey trivial checkbox in the material system or preview mode - but maybe we have to ask for a small project file for debugging (sent to our email).