I’ve taken to distinguishing between science(v), the method and science(n), the body of models and data. Science(v) is imperfect, but basically as close as we can get to objective truth. Science(n) can often stress conclusions further than their rigor justifies, but eventually regresses to the mean for the most part.

You can’t really question science(v) beyond its intrinsic epistemology, and no other method can really do any better. You can often question science(n), heck I can’t count the number of times “consensus” flip-flopped on red wine, coffee, fat, and so on. But eventually science(v) does bring science(n) to a stable empirical baseline.