I tried to do a deepdive on nutrition a few years back when I went on a diet. My conclusion was that there are far too many confounding factors for any population studies to be taken too seriously. That's why the field is so vogueish like you point out.

Long term, highly controlled trials are more useful, but much rarer.