A better idea might be to isolate all the various parts of the image and their influences on the other parts and then just compose an image adding these together. Then all you'd need would be to adjust the hue, saturation and brightness of each component and just add them to an image, instead of rendering the whole thing.
Yea, this is a better way to do it imo. Problem would be getting accurate colors via post tweaking, due to the custom rendered lighting in different scenes though.
An alternative would be rendering every color in SP's (for example) palette, using the same static scenes, isolating one, and then using it as the mask for all the other colored composites. No idea how long that would take
, but it could be a more accurate way of doing it.
A transparent background would also be most useful for photoshopping and use on a site, as you could composite any number of other backgrounds later.