So you could specify two or more images that overlay each other.
I believe GIF89a has many ways to handle this situation ranging
from just doing it, to timed delays (for simple animations) to
awaiting user input before continuing.
This would mostly solve the original problem. You'd still have to
generate a GIF with the overlays, but that really is little more
than catting the images together.
I'm sure that such a solution will not be highly considered by
either the HTML 3.0 people or serious image people, but it may
be just useful enough to add in. Go lobby your favorite browser
authors if you're keen on it. For transparent GIFs I had to
write the (trivial) Mosaic patches and then piss off Eric Bina
mightly to get it included, but it's there now.