I'm on record as not particularly liking the preproposal system that NSF Bio has been essentially forced to moved to. However, I will admit that some of the advantages of the system do indeed accomplish what they were set in place to do. Namely, the short format means there is a low activation energy for getting in the game.
Why is that important? For reasons outlined in Michael Tomasson's latest post about dealing with triaged proposals. Essentially, every PI needs to have a few ideas in the mix and getting feedback on them quickly is key. If your proposal gets trashed in review, then it's time to rethink it and move on with something else. The preproposal system allows you to do this, but providing feedback on a general idea (and in 4 pages, that's about all you can do) for relatively little effort.
The downside, of course, is that you have 8 months before you can resubmit. This is why it's important to be cycling several preproposals at the same time. With the 4 page format you can afford to do it, even if you're juggling it with teaching.
If you're tenure is largely dependent on bringing in NSF funding, you need to avoid being stuck with a single proposal you keep trying in slightly different ways and be willing to walk away from an idea that gets poor feedback at panel. By at least year three you should be maxing out the 2 preproposal limit for your favorite division and probably looking at ways to hit other calls as well. There are plenty of "cross-cutting" mechanisms to explore, as well as the CAREER awards (which can also be a landing spot for re-thought non-invited preproposals).
The most successful PIs I know have also been the most flexible in their funding search. Preliminary data can be used for multiple different proposals with different foci, as long as you can be creative in the questions you want to explore.