Just came across this (rather old) post.
I'm managing a team of dedicated frontenders and backenders. It'll be a cold day in Hell, when they start doing each other's work :-)
My concern with the community score, is that it gives a false velocity for the team, in the case of having dedicated Front- and Backend developers respectively.
EXAMPLE:
Say a story ends up at 20 story points, after a healthy discussion in Refinement.
Of these 20 points, it was clearly the backend developers stressing the complexity of building their part, whereas the frontenders needed only implement a new form.
So the backenders account for 80% of the story points (so to speak, they are the ones who raised the score).
Now, let's say the team has 4 backenders and 2 frontenders. And they make the above story in the sprint. All is well, and velocity could be said to be 20 story points per sprint.
Next refinement, tables are turned, frontend is a nightmare, backend a breeze, but it still adds up to 20 story points.
Given that frontenders now have to do 80% of a 20-point story, with only 2 frontenders to do it, how does that affect velocity?
For the sake of predictability and getting a trustworthy velocity, I think the community score has some major flaws. Unfortunately, I have yet to find the fix for them :-)