Hang on, so all the resurch we relie on because its published is in reality compleatly unverified? (except maybe cochrane which is meta resurch)
No. First of all, the person publishing the research has presumably done some work. If the research is experimental, then the results will be published, along with the details of the methods used so that other scientists can replicate the experiments themselves.
Not every piece of research is duplicated by somebody else. As a rule of thumb, the more significant the results (or groundbreaking), the more likely it is that others will jump on the same bandwagon and try to replicate and extend the previous research.
Theoretical publications are different. If a result is purely theoretical, then any glaring errors will usually be picked up by the expert referees - that is the point of peer review. Subtle errors may slip through initially, but once a published paper has been read by many other experts in the field, any errors will be pointed out by others in their own publications or letters to the editor.
The marvellous thing about science is that it has self-correcting mechanisms built into the process. Sooner or later, most errors are found out, and any fraud (which is rare) is uncovered.
Obviously, it is the truly unexpected and new results that tend to get the most attention from other scientists. Many papers publish something that is new but fairly obvious, and those papers may sink more or less without trace. As a general rule, the more citations a paper gets from other scientists, the more likely it is to be free of error - or else it is getting cited a lot because it is full of error and people are pointing out that fact.
Ahhh, I think this is a difference between physics and the other hard science fields in that physics seems to have a theoretical sub-field. So maybe what I said above isn’t exactly true for the cosmology and astrophysics fields. In the biological sciences there isn’t much purely theoretical work; it’s essentially all experimentally and observationally driven.
Many physics journals are unlikely to publish a mere proposal for an experiment, whereas the same experiment that has actually been carried out by the author(s) is much more likely to be published. Interesting results are more likely to be published than null results, and that's not always a good thing.
If you're publishing in an area such as string theory, there simply
are no experiments at this point, which is also not necessarily a good thing.
