So a lateral opening moving in steps from one end of the pouch to the other (no matter which direction) seems to fit best with natural selection.
I think you are making the mistake of assuming that either end opening configuration is the primitive state and that the alternative needs to therefore be superior to it.
If you look at the Dasyurids, often considered a more primitive clade of marsupials, many have a very rudimentary pouch which is little more than a circular fold of skin covering the teats (
Wooley et al., 2002). What we are seeing in such cases rather than an intermediate step between a backward and forward facing pouch is more likely the ancestral condition from which both of those states are derived.
It still seems weird, though, that a pouch that opened in the middle would be superior to one opening on the end, and give the animal an advantage.
If these were the first forms of pouch to arise then all such rudimentary pouches need to be is better than no pouch at all.
TTFN,
WK