"For a gene controlling alcoholism, it would have to appear somewhere where alcoholism had some substantial impact on survivability"
No, it doesn't. It needs to be a gene that has an impact on alcoholism, but the thing being selected for could be something else entirely. Consider sickle-cell anemia. If you just look at it as the "sickle-cell anemia gene" you'll boggle at how it spread as far as it did, but that's because it's also the "malaria-resistance gene". If alcohol is introduced into a population that did not have widespread, ready access to it for long time frames, there's no guessing in advance what could happen as it interacts with arbitrary other adaptations that have never "seen" alcohol before.
I want to point out you keep boggling at how intrinsically impossible this all is, and I keep citing concrete examples of very similar things that happen. And I'll reiterate that my point is that it was a viable theory, until the facts contradicted it, not that it was certainly correct and certainly not that it's correct now. Simply that it was not a stupid theory. (Indeed, given that this is just one study I'd be glib even saying it's "disproved", it's merely one that we must Bayesian-update to a much smaller probability. But I wouldn't downgrade it to zero quite yet. We've got a lot of gene science left to learn. And as always happens, when we take a closer look, it always turns out to be more complicated than we thought...)