I just made it use 4 active objects because I didn't want to spend time making a huge mod for testing. The performance is the same if you make it the max allowable, since I didn't optimize this code, and it's checking distances to everything. Since it's just checking distances to the centers of objects, the performance should scale okay to moderately large environments.
Making it precise, and work with big flat objects, etc., will slow down the code a lot. For rectangular prisms, I haven't been able to figure out a method to get real distance that doesn't involve at least one point rotation (9 multiplications) per object (and possibly a distance formula [3 multiplications] on top of that), at least every few frames.
I'm not totally clear how I want to approach it. The easiest way, and probably the most reasonable, is to just have modders put objects into groups of say 8 objects each, and they can specify a center or a bounding container, and track the positions of Tori and Uke, and just solidify the group closest to each of them. This will put the responsibility on the modder to specify things correctly, which I don't think I like.
Since the objects are static, I may end up writing code to parse, and analyze, and set up a set of buckets, and store closest objects in a big 3D matrix. So all of the computation happens all at once, and then it's just a matter of accessing a hash table. And I can write the table to a file, so next time it can just read the file, instead of redoing the calculations.
