Your web application, deployed as a managed instance group, is experiencing low memory on some compute instances due to a suspected JVM memory leak. You need to restart these instances to reclaim memory without dropping the available capacity below 80% during the process, and you aim to do this as quickly as possible. What is the best approach?