I personally really like the proposal here: http://research.microsoft.com/en-us/um/people/simonpj/Haskell/records.html The wiki doesn't show any opposition to this system. If Haskell had that record system now I would be very happy and would be fine with leaving other record systems as purely research until this whole research area comes to some decisions.
I believe the way forward is to implement several of the possible systems, and release them for feedback. To get users to actually try out the libraries, I think we need some concrete syntax for constant records, so I suggest we put in a feature request.
It would also be nice if one saintly person could spend the time documenting the available record systems in one document, trying out examples of codebases and collecting surveys on the various systems or something. It's a project in itself. Personally my vote for what it's worth is Worse is Better in this case, and to implement Simon's proposal (not that I think this proposal is Worse, but possibly worse than X-other-really-nice-but-tough-to-decide-on-system). If we still have nothing by this time in six months I'll implement the bloody thing in GHC myself.