An ECS service uses the rolling-update deployment controller. A new task revision repeatedly fails startup health checks, leaving the deployment unable to reach steady state. Operations wants failed deployments detected and rolled back to the last completed deployment. Which setting should it enable?

1. A task placement constraint that always selects the same Availability Zone.
2. The ECS deployment circuit breaker with rollback enabled.
3. Increase the service desired count and leave automatic rollback disabled.
4. Disable task health checks so the deployment is declared healthy regardless of application readiness.