Point Collision (Geometry nodes)

Poisson Disk is very effective in case of equally scaled objects or slightly random scaled. but when it comes to very obvious randomization I think we can make Minimum distance in Poisson disk as field value instead of just a plain float value , hence the minimum distance can be grabbed per point (according to scale instead of grabbing bounding boxes distance if it too memory consuming) by capture attribute node or even a dedicated node (Object collision node for example) of course it will better to get bounding boxes distance but it seems not memory efficient or it would have been applied since GN started instead of the traditional Poisson disk