Fred Kaplan's Take
Fred Kaplan, who offered up what I thought was an uncharacteristically blinkered and churlish dismissal of the first Lancet cluster-sample study of deaths in Iraq, has a more measured critical take on the newer study that raises two somewhat convincing points. One, as you may have seen on some other blogs is an argument developed by British academics Sean Gourley, Neil Johnson, and Michael Spagat who point out that the Lancet's household samples were located on "major commercial streets and avenues" rather than, say, back alleys. This significantly simplifies the logistics of doing the survey and for standard public health purposes works fine. For a war, though, it's at least plausible that the main streets might feature more violence than non-main ones.
Kaplan's other point is that the Hopkins account of Iraq's pre-war death rate (5.5 per 1,000 per year) is at odds with the UN's estimate (10 per 1,000 per year) and that if you take the UN's pre-war baseline, you wind up with more like 300,000 excess deaths than 650,000. I don't really know how to evaluate that dispute and I'm not sure anyone else does either. That said, this point from Kaplan seems to me like the crucial takeaway: "Let's say that the study is way off, off by a factor of 10 or five—in other words, that the right number isn't 655,000 but something between 65,500 and 131,000. That is still a ghastly number—a number that, apart from all other considerations, renders this war a monumental mistake."
He doesn't get explicit about this, but for my part the point would be that for a war allegedly justified at this point on what are largely humanitarian grounds to have any positive excess death rate is a scandal. A humanitarian intervention ought, on any reasonable view of the matter, save lives not increase the volume of death. The exact amount by which the death toll has gone up is sort of neither here nor there. On top of that, it's worth noting that there are other problems with the methodology that may be leading to an undercount. In particular, one assumes that in some instances entire households have been killed, but the Hopkins method isn't going to count any households like that.