Was thinking about collision detection a bit. Particularly looking at this article http://www.wildbunny.co.uk/blog/2011/04/20/collision-detection-for-dummies/ and the section(s) on minkowski sum/difference. The gist of this type of collision detection is rather than moving the objects and then detecting that two objects are colliding, we only move the objects as far as they'd be able to go before they'd collide. This is nice, as we can correctly calculate the time and position a collision occurs. If that time is less than total move timestep we can sim multiple times per timestep to correctly respond to multiple collisions per frame.
The trick (and what I'm having a hard time visualizing) is how to handle collisions of multiple objects moving in a single frame.
What I feel the correct thing to should do is calculate the first contact time within our timestep between all collideable objects, sim the objects to that point, resolve the collisions, rinse and repeat until we've exhausted all of the time in our