I'd say that the correlation between rate of aging and cleanliness of the electron transport chain (i.e. proportion of electrons that turn oxygen into superoxide rather than water) is the best evidence in favour. It's so general -- between birds and rodents, between CR and ad lib, between Peromyscus and Mus -- and I know no counterexamples. As for the MsrA report, it's about as good as a life-shortening result can be (in that when no extrinsic oxidative challenge was imposed it killed no mice before adulthood, so developmental problems can't be blamed), but -- of course -- it has the usual problem that shortening lifespan can be done in other ways than by accelerating the processes that normally kill the animal. Also, I'm somewhat concerned that the wild-type mice lived only 680 days average despite being C57Bl/6J. The MnSOD+/- mice are very interesting, but they only imply that the types of damage found to rise are not important in aging, not that _no_ damage protected against by MnSOD is important in aging. For example, Richardson's group has looked for various types of mtDNA damage but not for deletions, which are the lesion that's found in just about all OXPHOS-compromised cells. This is rather similar to the finding that CuZnSOD homozygous knockouts are basically OK -- it reinforced what we sort of knew, that oxidative damage in the mitochondria is likely to be much more important than elsewhere. (My money's on deletions being little changed in the MnSOD+/- mice versus wild type.) It's hard to identify -- or even to imagine -- evidence that seriously challenges the idea of free radical-induced damage being very important in determining the rate of aging: given the mountain of evidence in its favour, the simplest interpretation of any ostensibly negative result will always be to narrow the options for precisely which pathways from free radicals to aging are the most important, not to say that no such pathway exists.