I don't know a lot about this, apart from the fact that the implementation is a bit contentious. Can you elaborate on some of the opposition to refinements?
Brian Ford, one of the heads of Rubinius, has also been very outspoken about refinements, though I can't remember a specific link at this time.