Question about valid AABBs

(Considering here within a single BLAS…)

Is it valid for AABBs to overlap within a single BLAS and should all (vendor-specific) acceleration structure processing correctly detect hits/entries into each?

For example suppose I have implicit spheres inside tight AABBs, and the spheres partially overlap, such that the centers of the two spheres are closer than the smaller radius of either sphere, and thus the AABBs also overlap/intersect for a non-zero volume region.

And related question, is it valid for one AABB be fully nested within another AABB?


I think overlapping is a distinct feature of BVH. Otherwise one might use a different space partitioning data structure.