I have a problem. If I fracture a mesh, I have to leave a visible gap between meshes so that the physics engine won’t immediately explode them away from each other at a million miles per hour, this isn’t ideal, because the cracks are visible prior to any collisions. I know this can be overcome by hiding the fractured geometry until the point of impact and showing the unfractured geo up until that point instead, but even then when the fractured objects are made visible again on impact, the result is very unrealistic because the collision may not move all of the fractured objects, but they’ll still have a visible gap between them.
I think a good solution would be if internally collision meshes were smaller than the actual geometry (based on the collision margin). This would mean that the collisions would react properly, the cracks wouldn’t be visible, and there’d be no need to hide groups of fractured objects prior to collisions.