We give complete mathematial foundations for the method, recently developed by Chou, Mannava, and Park, for verifying safety properties of cache coherence protocols. The method employs a specific form of counterexample-guided abstraction refinement and is originally described on worked-out examples of the German and FLASH protocols. We describe and prove the method at an abstract level, thus es...