An interesting thought indeed, however, I am still not convinced it will look like a normal sine scroller. Any of the 16 slices can have any y-position, e.g. col1 = y, col2=y+(whatever). As far as I can see Mrs. Beanbag's approach will give a somewhat static scroller but yes, it depends on the used masks. Also, I'm not sure if it will be much faster since it requires one additional blitter source = more DMA used = slow down.
It would be interesting to see an example of this approach though so it can be compared with the speed of a highly optimised sine scroller.