The current implementation is different from R3F, as you can see, the model is treat as a group (I personally prefer this approach) is best performance and cover most of the cases, BUT, not all for some specific cases maybe this is not a viable solution.
R3F handling the models in which each geometry has a different RigidBody/Collider:
I don't know if this will create future problems for uses cases or the people can workaround them. I don't know if this makes joints cases when models more difficult.
Suggested solution
Proper discussion and research
Alternative
Just by the record, I like the current approach (but maybe I'm thinking in just low poly games)
Description
The current implementation is different from R3F, as you can see, the model is treat as a group (I personally prefer this approach) is best performance and cover most of the cases, BUT, not all for some specific cases maybe this is not a viable solution.
R3F handling the models in which each geometry has a different RigidBody/Collider:
I don't know if this will create future problems for uses cases or the people can workaround them. I don't know if this makes joints cases when models more difficult.
Suggested solution
Proper discussion and research
Alternative
Just by the record, I like the current approach (but maybe I'm thinking in just low poly games)
Additional context
No response
Validations