In models where outcome probabilities depend on distant outcomes (not just settings), signaling cons...