If it can scale in low-res mode, I wonder if that's how the burning force light effects are done. (ie software rendering to a low-res translucent bitmap scaled up 2x -doing it low res on the VDP1 end wouldn't make much sense since you'd still need to draw the same number of pixels either way, but doing that on a separate VDP2 bitmap would make sense -less CPU overhead and less RAM needed)
And another more general note:
For blitter rendering, using a painter's algorithm for rendering (be it 2D or 3D) would be particulary attractive for chunky-pixel renderers rendering few (or especially 1) pixel per write and especially without a z-buffer or similar pixel priority information (for 2D games, much lower res depth could be useful where a 3D z-buffer would be worthless).
It would seem like the Saturn would be particularly favorable for this with VDP1 rendering 1 pixel at a time, and using software Z-Sorting (ie on a per-sprite/primitive or per-model basis) since the Z-sorting could be used to directly organize objects to be drawn with a painter's algorithm. (so you'd just be doing writes to the framebuffer, no read-modify-writes . . . unless you used translucency effects)
On systems rendering phrases of pixels at a time, it would depend more on hardware features to cater as heavily to painter's algorithm type rendering. (a 2-pixel wide bus -like 3DO or PSX- should still get away with only doing writes, but would need logic to support that -either writing a 32-bit word if both destination pixels are opaque, or a single 16-bit write if 1 pixel is transparent, and a 4-pixel wide bus would be more complex still -using single writes in cases of 1/2/4 consecutive opaque pixels, but a read-modify-write in the case of mixed opaque/transparent pixels -again, requiring logic to support that, otherwise requiring a read-modify-write for all rendering)