AS periodically checks the health of instances in the scaling group and identifies all instances that are in poor health. After the AS marks an instance as unhealthy, it schedules the instance for replacement.