it is not who you know...
Of all the agencies, the NSF really takes any conflict of interest, or appearance of conflict of interest, the most seriously. In particular, all proposers must list the PhD advisor and postdoc employers, and all recent (last 5 year) collaborators and co-authors. Obviously all blood relatives, relations by marriage and institutional colleagues are also barred from refereeing proposals (though I have my doubts about the last one, institutional colleagues are not all that often familiar with each others research and I don't think they are particularly likely to be kindly disposed... just kidding! The conflict there is financial, institutions benefit collectively from grant awards, I get a grant, my department and college benefit directly...)
So, this works fine for Ye Olde Pundit mode of doing research, where Professors are Professors and All Others are Mere minions. Except possibly for one, or two at most, equally disinguished colleagues and Dear Old Collaborators. After all, some fine research has been published with two senior authors on the paper.
But, a lot of modern research involves Big Collaborations. DoE PIs have hundreds of collaborators and co-Is, but then do we really want them to apply for NSF funding? I don't think so.
But, cosmologists are kinda cool, and they too have MegaCollaborations with O(100) co-authors on some papers. What to do?
And, this collaboration stuff is insidious: eg, consider a medium sized project with a Result to be published. Let us say it involves actual observations, at more than one wavelenght or observatory (with, of course, a decent amount of theorising mixed in). I, and my co-author Prof A, may want to invite our colleague Prof B to participate, since Prof B has a particular expertise or unique observatory access. Prof B, in writing his bit, gives some data to Student C, who is well along and is set up to correctly and quickly reduce just such data (but is not directly working in this sub-field). C then contributes a figure, a methods paragraph and eliminates the stoopid conjecture Prof Z once came up with to explain these things.
So now student C is my co-author, I have never met her, and only know that Prof B who is a friend of Prof A vouched for her competence and she did a minor but important sub-section of the whole work. Several years ago.
Now, 5 years later, student C is hot-shot prize postdoc C about to become tenure-track Prof C, and is certain to be on the NSF panel. But may not review my proposal...
Do a few of these; and the only people who can referee your proposal are either those who do not work in your sub-field, or those who are in direct head-to-head competition with you.
Fortunately, these people are both competent and have no identifiable conflicts of interest.
For extra refinement: the NSF wants the current institutional affiliation of your dear collaborators.
Now, you tell me J. Smith (London in 2000); H. Lee (UC in 2000) and H. Muller (Euro StadtUniversitat in 2000) - where are they now? might they be on an NSF panel if you miss them? and could they ever tell?
Now repeat 100 times. And the proposal is due in 97 minutes...
The Good News is that NASA's new on-line proposal system seems to have been rough enough to have depressed the actual number of valid proposals to make it through the system.
NSF's Fastlane does not have this problem, it works, people know it, and it is good to use. Fear not, the NSF will have to transit with the other agencies to Microsoft oriented grants.gov where every PDF file will be corrupted and no actual direct PI access may be granted, so someone else will be entering you grant info into the system. Piecewise. In doc format.
Maybe the Federal Budget deficit can be cut after all.
On a different line, a small individual PI NSF proposal is 30-40 pages. of which typically 15 are the actual proposal, the rest are auxillary and supporting documentation. Having all electronic submission is very important.
NASA proposals are longer, even though individual unsolicited proposals are restricted to 15 pages - last I checked the new NASA cover "page" by itself was 18 pages.
With paper submission (through this year), NASA wants 15+1 copies. At 50+ pages per copy. Far as I can tell as few as 3 and no more than 7 or 8 of those copies are ever read by a human being. Include institutional archive copies and you can get to 1000 printed pages per proposal. Or 50-100,000 pages of paper per sub-panel for unsolicited proposals...
Observing proposals, thankfully, tend to be shorter. Solicited proposals may be shorter, as little as 1 page, or longer, as much as 100+ pages per copy.
And who says the Federal government is not combatting global warming? They have one of the largest carbon repositories in the world buried in shallow concrete storage in DC (ready for quick retrieval to stop the ice age, or something).
Hm, a proposal to convert archival federal paperwork into high octane fuel for internal combustion engines... Hmmmm.