The idea looks to me to have two distinct elements. Figuring out what people use (where the need is) and then figuring out what there is that people use that needs work.
I like reuse. I suggest popcon for the first element and replicating as much as possible from Debian's WNPP for the second. The first should generate a list indicating what is most used and the second would crowdsource the ability to identify where work is needed.
The way popcon could be used is to run it extensively against Fremantle users and against Diablo or Mer and then diffing. That should give us a clear indication of what is popular in Maemo but not available (yet) in Fremantle.
Debian's WNPP is a good choice for Maemo as the semantics and procedures are familiar and - being well established - quite thorough and well thought out.
Caveat: As the N900 becomes more generally available I expect to see a lot of packages organically being ported which might obviate the need for this kind of solution.