This has come up before, but I heard it again in conversation with a PO this week: The average length of an NSF award is going up.
The NSF core programs have always had the ability to award anything from 1-5 years in duration. For whatever reason, PIs have settled on 3 years as the "normal" time for an NSF award. I don't know the history here, I just know that when I first started applying everyone told me not to do anything (on the non-scientific side) that would stand out to reviewers. This is not unlike NIH n00bs being told to stay with the modular budget.
Three years wasn't so bad when funding rates were higher, but these days it means that you really can't take a break on submitting if you want to keep the lab continuously funded. Yeah, you can use no-cost extensions to effectively make a grant last 4 years, but it seems that people are just starting to ask for 4.
File that away for all of you who are waiting to hear back on preproposals.