It might not be that fast if implemented via blueprints but I can’t think of a reason why it would not be possible.
You can find multiple articles and guides about quadtrees in google if you want to research this further.
I would start by creating 6 quadtree grids, spherify them, offset them via noise etc. It’s the same thing as you are currently doing really, it’s just how you generate the initial grids that’s the tricky part.
The nice thing with quadtrees is that you can skip generating any grids outside the camera frustum and “subdivide” the grids based on pixel coverage so the vertex wastage is pretty much as little as possible, it’s very efficient.