Alexander Fleming's original plate, containing penicillin-producing mould. Fleming's accidental discovery was ignored by the scientific community for years.
There have been countless flukes in science. To name but two: the animal blindly selected by Axel Holst and Theodor Frolich to study scurvy was fortuitously one of the very few species of mammal that cannot produce their own vitamin C (guinea pig); while Luis Pasteur discovered virus attenuation as a method of vaccination creation because his lab assistant left virus broth out for longer than intended, inadvertently vaccinating chickens against cholera instead of infecting them. And then there was this:
Hence much of scientific progress is down to chance. So should we trust it? The answer depends on how much faith, so to speak, we think we should place in the scientific method. For how do we really know that we land on objective truths when our results only statistically correlate with our hypothesised theories? In such a case we can only favour one possibility over another (nb. ‘underdetermination’ in philosophy of science).
Perhaps, as time passes, as scientific knowledge is accumulated in one great pyramid of findings, we only construct what Foucault calls the constitutive limits of discourse.
Notwithstanding these concerns, many scientific realists (specifically, ‘structural realists’) believe that, even if we never have concrete facts about the physical world, it would be a miracle if we weren’t accurately describing the Universe’s fundamental structure to at least some degree (n.b. ‘no-miracles argument’).