While I certainly agree with much of the factual content presented both in the article and in the comments, I think that science already has a lot of self-correction mechanisms built in. None are perfect individually, but the big, messy system has a lot of redundancy built in. It's just not always so visible to journalists or science writers, who don't hang around the scene for the years that it often takes for science to find its way again, so to speak.
For example, many of these high-profile, possibly erroneous (or occasionally fraudulent, it seems) Nature or Science articles are high-profile because they seek to address a contentious or long-standing problem in the field. When this happens, there are typically existing alternate hypotheses. It's much easier to get papers published or grants funded by seeking to test competing hypotheses than to simply try to verify an isolated study. It can also be easier to find weaknesses in an individual study by testing it in a different way, or against other models, or whatever, than by simply trying to reproduce it. Often, a single study might be impossible to directly replicate, or the underlying flaws may not be apparent until the problem is approached from a different angle.
Granted, this can take a couple years or even decades, but falsehoods (intentional or not) tend to become more apparent as their context becomes more clear.
For example, many of these high-profile, possibly erroneous (or occasionally fraudulent, it seems) Nature or Science articles are high-profile because they seek to address a contentious or long-standing problem in the field. When this happens, there are typically existing alternate hypotheses. It's much easier to get papers published or grants funded by seeking to test competing hypotheses than to simply try to verify an isolated study. It can also be easier to find weaknesses in an individual study by testing it in a different way, or against other models, or whatever, than by simply trying to reproduce it. Often, a single study might be impossible to directly replicate, or the underlying flaws may not be apparent until the problem is approached from a different angle.
Granted, this can take a couple years or even decades, but falsehoods (intentional or not) tend to become more apparent as their context becomes more clear.