I keep coming back to what Nick Lane put in a footnote in his spectacularly good Transformer:

It is worth reflecting that paper after paper published in top journals can be essentially wrong. Please don’t think that these papers were especially bad. Far from it. Science is hard and we are all wrong a lot. The harder the problem, the more wrongness there will be. If you want to take a moral away from this, then assume that most published papers are at least partly wrong. That helps to explain why scientists squabble a lot in public. But these arguments are part of the way in which science corrects itself. Publishing detailed papers lays out exactly where scientists are wrong, and how to ask better questions. Unlike any other human endeavour, to my knowledge, the scientific method is a ratchet for improving answers over time. Being wrong is part of the ratchet, just as detrimental mutations are part of natural selection, driving the evolution of life’s wonders.