Move a Skeletal Mesh Collision, but move the visual mesh seperately?

Is it possible to separate a Skeletal Mesh (or even a Static Mesh) collision from the visual mesh, and move that separately of the actual visual mesh? Of course, this will no doubt have to be done in C++.

The reason I ask is because after reading up on the Character Movement Component, I see that in order to maintain server-side accuracy, the Collision Capsule of a character (well, the actual object itself since it’s the root) is moved when an update is received, but the actual mesh itself ‘Blends’ to that location over time. The idea of this I guess is so that Collisions remain as accurate as possible and the local clients don’t get a hard ‘pop’ when they receive updates. Nice way to do it!

However, I have my own Pawn class which has to use the Collision of the Skeletal Mesh, since the shape and size of the meshes can vary so much, using one of the collision shapes isn’t really possible for me. The best solution I can think of would be to create a ‘Custom Collision Shape’ component and be able to tell it to use the collision I’ve generated on the Skeletal Mesh. The only other alternative really is using a Static mesh or something as the collision, which would be horribly inefficient.

So basically, is there an existing ‘Custom Shape’ collision component I can use, or if not, how would I start going about it? Realistically, the mesh and collision need to be separate components.