Now that we have the ability stop/start existing instances, another nice addition to the instance cap behavior is to have an option to count stopped instances as well as running instances. This would allow for forcing a fixed number of instances to be created, ensuring that a job is always run on an instance with a warm workspace, or queued. In my case, starting a job on a new instance is usually the wrong thing to do because a cold workspace increases by build time by 5 times or more, waiting for a running build to finish is a much better choice.
Now that we have the ability stop/start existing instances, another nice addition to the instance cap behavior is to have an option to count stopped instances as well as running instances. This would allow for forcing a fixed number of instances to be created, ensuring that a job is always run on an instance with a warm workspace, or queued. In my case, starting a job on a new instance is usually the wrong thing to do because a cold workspace increases by build time by 5 times or more, waiting for a running build to finish is a much better choice.