What to list as affiliation when work spans multiple institutions?

Most journals understand that such complexities exist, and will allow you to list multiple institutions, so all four of your options are certainly viable.

That said, given that you have done much of your work on the paper while at your new institution, your new institution should certainly be listed as an affiliation, which means that you should be using either your option #2 (new only) or #3 (both institutions).

That leaves the question of whether the old institution should be listed or not. From what I have seen, the customs on this tend to vary by field and also by institution: you can probably make a reasonable argument for either listing or not listing the old institution. I would thus recommend talking with your co-authors to see what their feelings on the matter are, particularly if you have some co-authors from the old institution on the paper, who are the ones most likely to be upset if your judgement disagrees with theirs (in either direction).

I'd list "current affiliation", and acknowledge the other institutions where you worked. If some of your coauthors still reside there, you might just leave that out (the institutions are mentioned as their affiliations).