Couple thoughts on this - for what they are worth:
Would undo levels have anything to do with this?
Also it sounds like interaction of light sources on the objects is imperitive gauging from the descriptions above. However, if this is not the case, one big breakthrough that occurred to me awhile back is that you have a virtual movie studio with your computer complete with motion control. This enables you to perform several "less complicated" passes that can be composited later in Quicktime Pro.
However, such a solution is not ideal as of yet (at least with Renderworks 9) as no alpha channel can be made requiring the user to use color mattes - which gets tricky. (thumbs up for alpha channels in future versions!)
Granted, only a possible workaround in the least, but thought I'd throw that out.
Continued luck in solving it!