>"progress"
I've determined that I'm going to have to delete my existing AABB collision code; not trying to stuff OB collisions into my existing funky algorithm makes everything way more tractable. But it is upsetting to lose polished code. Also the fully optimized code from the dude's paper isn't working, there's some transpose/row-major nonsense going on but it is not being consistent... I should just use the full dot products... "make it run; make it right; make it fast".