The NSF preproposal is a fairly new document and people who submit to DEB and IOS are still trying to figure it out. I'm currently reading proposals for my second preproposal panel and some patterns are starting to emerge. In particular, how people handle the Specific Aims section 1) makes a big difference in the flow of the document, and 2) is pretty heavily correlated with those I suspect have NIH experience.
There seem to be three flavors of SA that I see re-occurring:
1. Just the facts
Some PIs are simply stating the aims of the project with no supporting text. Just two or three aims at the top of the document and then we'll tell you about the background.
2. The hybrid
Others are including a bit more than approach 1, but wrapping the whole thing up in half a page or less. There's some context, but it is mostly focused of fleshing out the Aims a bit.
3. The pager
These generally stick to the NIH format of taking a page to nail down wtf you are proposing to do. There's a certain format to these pages (one opinion here) and it should do a good job of summarizing the science of the proposal concisely.
I may be biased here, so take my opinion with a rock of salt, but I find option 3 to be far more readable a format. Option one is jarringly disjunct and option 2 never seems to do quite enough for my reviewing tastes. My guess is that some people feel that the third type of SA section is redundant with the project summary, but the summary is very specific and includes Broader Impacts. If you're sly, you can use the summary to include a few interesting tidbits before hammering your best stuff home in the SA page.
Like many novel NSF documents (see: postdoc mentoring plan, data management plan), it's going to take a few rounds before things settle in and people have a feel for what to expect. For those of you who have written preproposals, how did you handle the SA section? What are you reviewers seeing?