In my view the most practical solution would be to start with a clear reputation system for reviewers. Editors already keep remarks and statistics of how referees perform but this information is not public.
Peer reviewing is a very important part of the scientific process and it should have a higher value. Without any way to keep track of who is performing this vital role this can never be included in an evaluation system.
One nice side effect would be that journals could find new referees from this common pool of public evaluations.