Decentralized consensus works in Bitcoin (a minority can never fool the majority) because verifying work is much easier than doing the work. Decentralized consensus doesn't work very well in research, because verifying work is even harder. If you repeat someone's experiment and get different results, maybe you just didn't do it correctly? So you need multiple replication attempts before you can be sure to have falsified it, and in that time someone else has already published the next bogus study.
The real question is why do governments watch private interests take over public research institutions and not only do nothing to preserve the lack of bias, but even go out of their way to help the process? (And yes, the answer is not secret.)
How government money gets spent is a very political process, and naturally that puts a lot of pressure on research organizations to produce the politically correct result.
One of the troubles is human bodies are quite complex and able to thrive on a wide variety of diets so any simple looking at the correlation of factor A and result B is likely to be shaky due to all the other stuff that happens. One example I found amusing is the correlation between alcohol intake and health. My recollection is the UK government did some study to do health recommendations on how much to drink but found the healthiest drank something like 4 pints a day so thought we can't say that and recommended max 2 pints or some such but the thing is cause and effect are basically the wrong way around - people drink till they feel ill and the healthy can drink more. There's a lot like that.
My rule of thumb is to ignore the 'science' stuff unless they can quantify how many years the bad habit takes off your life. Salt was never like that. The big ones are smoking - 10 years or so, obesity - of the order of 10, air pollution - of the order of 2 years.
You'd think that science would make things here a little more conclusive by now...