I think we lack a "let a thousand flowers bloom" approach here. Just create a way to fund lots of individual researchers and teams with small grants to study whatever they find interesting, with no crazy deadlines, credential requirements, publish-or-perish demands, administrative overhead, etc. It doesn't matter if 999/1000 just produce garbage, if the one remaining hits something big. I suspect modern science is like a gold rush, with lots of people flocking around a few fad areas, and o lot of the search space uninhabited, because if you stray out of the fad-topics, you get no advisor, no funding, no lab, nothing.
