Should Collision Geometry Nodes Be Flattened?

I was reading this from the panda manual:

Set this true to allow NodePath::flatten_medium() and flatten_strong() to combine multiple CollisionNodes into a single CollisionNode--but only if they share the same name and collide masks. When false, CollisionNodes are never combined. This is false by default, since collision tests rely heavily on bounding volume tests to be efficient, and combining CollisionNodes is likely to merge bounding volumes inappropriately. 

Right now, what I’m doing is calling flattenStrong() on all my collison nodes (most are collision geometry) and attaching them to the object they represent (which is also given flattenStrong())

Is calling flattenStrong() on collision objects that use collision geometry a bad thing?

Flatten strong makes objects be a single object. If you have a boat made out of parts and want to flatten everything you can - just call that collision or not.

Yeah, I was calling flattenStrong() to combine geometry.

The Question was relating to rather or not it’s safe to ‘flatten’ geometry that will be used for collision checking (such as collision geometry).

I’ve experience both ups and downs from doing this. All in all…Panda’s collision system is very expensive…even if you’re creating very cheap collisions and loading those collision in and out of the collision loop based on rather they’re needed or not.

The collision system will knock off some frame rate regardless of what you do, unless you’re creating a game where everything can have a box or sphere collision.