Optimizing through the physics engine?

I''m in the process of designing a new educational graphics engine (i.e. a teaching project of my own) and thought of a new idea I''m not sure has been tried before (possibly for good reason but still). The idea is basically to use the physics engine as part of the optimization of the rendering order - break up the scene into the possibly colliding pairs using the sphere test and then setup groups of objects which might interact. After the physics for one set of these is processed, they then get rendered and so the physics for the next set can be performed while they''re rendering. Would this result in a performance boost or would not enough polygons be being rendered to make it worth while? The engine is being designed to render planetary terrain and objects placed on it and above it. ---- flying starfish