Bad economics is the driver of most of these problems. When incentives are set up so that if you fudge results, you're more likely to get funded, then fudging results is going to happen. A lot. Hence the "reproducibility" problem.

Twice in a row I've had grants that got pretty good scores on round one only to be killed on round two (after, you know, "fixing" the grant they way they suggest). In my latest run in with the NIH, our scores from two of three reviewers got markedly better on round two, and the third guy gave us all 7s, after more like 4s and 5s on round 1. No explanation necessary. No recourse. That's just NIH for you. Peer review is a scam.

I were to get a stab at fixing NIH's funding problem, it would be to do what they did in Nigeria with their entrepreneurship grants. Namely, take the top 50% and put them in a lottery. No more study sections. No more grants only going to people who already have grants. No more fretting about getting another grant the minute you get the first one funded. Just luck of the draw if you meet the basic requirements. Easy.