Accelerated pairings - what do we want?

The FIDE Swiss Pairings Commission is having another stab at  defining a good method of accelerated swiss pairings, although it has already run into some predictable problems. The first is one of defining what a 'good' system is while the second is find a method.
For most people a 'good' system is one that minimises 'junk' rounds, although how this is measured can be tricky. Reducing the average difference of ratings in the early rounds seems to be a good measure, although this may cause unintended side effects. Another measure might be getting approximately the same standings in a N-2 round event as you would in an N round event. And coming at it from a slightly different direction, increasing the number of norms in the event might also be a goal.
As for the method a number of different systems are being trialled. The somewhat discredited bonus point system is being looked at, although the issue of 'bad' pairings when the acceleration stops is already apparent. Graduated bonus points systems (eg 1 bonus for the top half for 2 rounds, dropping to 0.5 for the next 2) are also being trialled, although the results aren't conclusive. Slightly more complicated systems are also being investigated, but for know it is still a work in progress.