Evidence that granting agencies should fund people, not projects

At least, they should if they want to maximize their chances of funding truly groundbreaking research, and are willing to accept an increased risk of failure. That's the conclusion of what seems to be a quite rigorous and detailed study of the work of HHMI investigators (HHMI funds people) vs. otherwise-similar NIH investigators (NIH funds projects).

I'd be very interested in a similar study of NSERC, which like HHMI funds people rather than projects, but unlike HHMI funds a relatively large number of people with relatively modest grants. In a sense, NSERC is 'hedging its bets' by funding so many people; HHMI takes the riskier approach of putting its eggs in fewer baskets. (NIH also puts its eggs in relatively few baskets, but because those baskets are individual projects that are selected for their high likelihood of success, NIH's approach is ultimately quite risk-averse) In funding people rather than projects, but funding many people rather than few, is NSERC thereby gaining the 'best of both worlds'? Or the worst? Or somewhere in between?