Yes, it was back and forth many years whether the dip at the bottom (with very modest drinkers being slightly healthier than non-drinkers) was genuine, or wheth

Yes, it was back and forth many years whether the dip at the bottom (with very modest drinkers being slightly healthier than non-drinkers) was genuine, or whether there was some remaining effect which hadn't been appropriately controlled for. But that was pretty much settled with Mendelian Randomization studies on exactly this gene. As the article says, some people have different versions of the alcohol dehydrogenase enzyme, which basically gives them a worse hangover, faster. People with this condition drink slightly less than they otherwise would, wherever they are on the consumption curve. Knowing that, and knowing that the gene is more or less randomly distributed in the population under study, and assuming the gene affects nothing else (a safe bet, since alcohol dehydrogenase is a fairly task-specific enzyme), we can disentangle the effect of the alcohol from all the other things that could cause you to drink more or less alcohol. Now the consensus seems pretty solidly on that the apparent protective effect of moderate alcohol consumption on stroke, cancer or heart disease is non-causal.