Hi Ivan,
No, we use AABB trees since that fits most scenarios pretty well and allows us to focus our performance efforts into one implementation.
I’d still ensure that this is the weak link before rewriting an entire broadphase though - this is a major component of the physics engine, and this still could end up less performant that the AABB at the end of it.
Obviously if you do though, we’d love to hear about the results!
Best
Geoff