But this avoids the central issue - how can we know if a paper helps people make progress on solving problems? Citation counts are just a proxy for exactly that - to move away from them you have to come up with a better proxy or evaluation method.
I think this exposes the bigger problem of metrics in general. It's hard to determine how much impact research has, particularly in the short term. I'd be interested to see what the outcome of the REF is, but I remain sceptical and apathetic towards it, as does everyone else I've spoken to.
[1] : http://www.ref.ac.uk/
[2] : I get the impression it does anyway - I'm not involved in it this round.