In thinking about this more, I created a new proposal, #131 (https://github.com/ghc-proposals/ghc-proposals/pull/131), which is the best alternative I can think of to this proposal. Indeed, I think it's the "more comprehensive solution" Iavor seeks. (Though it doesn't use nominal arguments.) Sadly, I don't think #131 is implementable, though it's specification is wonderfully straightforward.

I post #131 essentially as a counterpoint to the current proposal. If #131 were implementable, I would favor that and abandon this one.

Richard