I've implemented an octree to divide 3D space that contains several objects. I've noticed in certain octree nodes I could use more refinement in some axis, but splitting the node into 8 children is a bit overkill. Imagine a neighborhood full of buildings. At some point, it doesn't make sense in my case to further subdivide in height, but I need more precision in the other two axis.
Is there any major drawback to subdivide an octree cell into just 4 children, ie., subdividing in only two axis? This would create a hybrid between an octree and a quadtree.
I haven't found any documentation related to this, which seems strange to me, given it looks like a simple approach to a common problem.