Retries are great for transient failures, but they can't solve eventual consistency. Treating one problem as the other often adds unnecessary complexity to distributed systems.