Responsibility for checking capacity lies with Jenkins itself. It is the Jenkins core code that polls to check capacity. Once Jenkins has identified a need for more VMs, then it becomes the vSphere plugin's responsibility, and the time it takes for a new slave VM to come online is mostly dependent on how quickly the VM boots up and connects to Jenkins.
I've not measured the time it takes but I believe that more recent versions of Jenkins poll more frequently than older versions.
I would suggest that you check the logs for "vsphere-cloud" and "org.jenkinsci.plugins.vSphereCloudSlaveTemplate" and you should then see (at log level INFO) logs telling you of when the vSphere plugin has been asked to create new VMs and its progress doing so.
I would also recommend ensuring that you have the VMware tools installed on your VMs and tell the vSphere plugin to wait for the VMware tools to be detected on the VM before proceeding with connection BUT not to wait for long after that before connecting - this may improve the speed with which your VMs come "online".
Believed to be a core Jenkins issue; no response from original poster.
Feel free to re-open if there's new evidence.