In general, the more modularized you can make your "politically inconvenient" technology parts the better. The trouble with "but with a radio integrated!" projects is that when you change any other parts on the board, you are required to re-certify the entire thing, even if the radio portion hasn't changed. This spells trouble for fast-moving "versioned" parts like ARM cpus, especially when you're small scaled and the version you're on might get discontinued next month.
Atmel has been really good about supporting CPU footprints over many generations (their customers don't like redesigns either), and the FCC rules seem to give you a pass as long as you haven't changed things electrically. There's actually a reasonable amount of wiggle room - basically if there is no reason a board should perform differently, they say it doesn't need to be recertified. (at least, that's what they say)
That said, our goal is to brave the waters of the FCC to provide quality radio products to the open source community. If we succeed in getting funded, we should be able to handle the occasional new testing needed for any product changes.
We have some higher volume plans which will help pay for that. Arduino is only part of what we're doing. :)
This all started because there weren't any open source module designs for the CC1100 radio, so I made one and put it on github (actually I was using the 2.4ghz CC2500 at first). Eventually I slapped an arduino on, and boom, useful thing!
The chips broadcast right in the 70cm ham band (440 MHz). If you're a licensed ham, then you don't need FCC certification and can just build one yourself :-).
All the embedded projects I ever did with radiomodems on them (quite a few, including a laser game and associated hardware) had the radiomodems on them as daughter boards for that exact reason.
