1: Whatever the original alignment of the atom was, the N and S components experience a different lateral force, so the amplitude blobs corresponding to those two cases will be shoved apart sideways. I guess you could say they begin aligning the moment the blobs become distinguishable - and that'll depend on a variety of other factors like how you're measuring the deflection. This is the case whether you are working in MWI or Copenhagen.

2: back to unaligned, if you can completely reverse the forces in all cases and equalize the path-lengths - though the way the S-G apparatus really works, that's not going to happen. So you're going to end up with a complicated mishmash of altered phases, and thus reorientations.

3: The atoms heading through the hole in the first screen would end up with two single-slit interference patterns - though since atoms are really heavy I wouldn't count on this being a very noticeable interference pattern. It's two single-slit patterns because the left and right contributions are not feeding into the same quantum state (different spin), so they can't interfere.

4: By measuring the deflection, you've provided an earlier mechanism for distinguishing the blobs, so it would be at that earlier time.