I realized the cluster was using the maximum number of worker nodes (NodeAutoScalingGroupMaxSize) specified in the CloudFormation template. After contacting support they informed me there is a setting in the template to control the desired capacity (DesiredCapacity).