I came across either a problem or a bug. Hip file included. If you use an sdf as a collision for vellum, at some point higher resolutions break and the bounding of the vdb acts as the collision instead of the correct shape. Check out my video/photo. On the right the only difference is the vdb is higher resolution. Am I doing something wrong? I am using an sdf because the shot (this is a test scene) I am working on the collision has changing topology which vellum doesn't like. The problem is I need by vdb high res to capture all the geometry, but then it only works at low resolutions. What is happening here?