I should know these, and I have my own convictions/assumptions on these matters, but I just wanted to check with you guys, all-knowing wizards of Blender.
In terms of collision mesh, is it faster to set it to a solid shape, like, say, a box or sphere, than just to leave “Collision Bounds” disabled (for static meshes, of course, in which case Blender seems to create a triangle mesh, though suspiciously without the performance hit… From my experience, anyways)?
Does enabling “actor” impact performance, at least when an object is not being detected by some sensor (radar, etc.)?
Were I to create a large static environment procedurally in-game, would it be better to leave all physics meshes to themselves (without parents), or to coalesce their collision meshes into one “compound” one by parenting them to a central object with “compound” collision bounds enabled?
It makes no difference with Static objects. I tested this myself a while ago.
Edit: My suggestion would be, with generation of Static objects, use very low poly collision meshes. For example, initially you can spawn in the low poly mesh, and then replace the display mesh with a higher poly mesh. Or, you could even parent the low poly physics mesh to the higher poly (no collision) display mesh. All depends on what you are generating of course.