Rethinking Individual Global Max in Cooperative Multi-Agent Reinforcement Learning