Yes, stability is the big issue. My solution is almost certainly wrong, although instructive :-) The unit cube solution also seems unstable, but in a different sense: a tiny change in payoffs can grow the Pareto frontier a lot. At this point I'm wondering whether there is any continuous solution.