Christian Frohlich at the University of Tubingen, Germany, shared with me his paper with the above title. I’ve only skimmed it so far but this seems like an interesting alternative to the way I’ve been trying to infuse imprecise probabilities with statistical/frequentist properties.