A development team deploys a customer-facing web application to AWS Elastic Beanstalk running on a fleet of EC2 instances. Their priority is to avoid any reduction in application capacity during deployments and to guarantee that a failed deployment can be rolled back quickly without leaving instances in a mixed or degraded state. They accept longer deployment times as a trade-off. Which Elastic Beanstalk deployment policy best meets these requirements?