Absent from the many analyses and discussions of scientific peer review are two intangible but very important byproducts: 1) feedback to the applicant and 2) exposure of the reviewers to new hypotheses, techniques, and approaches. Both of these phenomena have a virtual mentoring effect that helps move science forward. Such learning can occur as a consequence of both manuscript review and grant application review, but the review of grant applications, by its very nature, is more iterative and impacts the direction in which research moves very early in the investigation.
There are at least two funding agencies that recognize this phenomenon in the actual design of their peer review processes, so they deserve mention. The idea is to include non-academics as peer reviewers precisely to effect the sort of co-production of knowledge the article above suggests.
The second is the US Congressionally Directed Medical Research Program. Details of their peer review process are available on their website here.