Settling the Reward Hypothesis