hudson.plugins.ec2.EC2Cloud.provision has calculates the excessWorkload in a way that doesn't take the ami configurations in account.

Because of that, if I configure two diffent AMI's:

AMI#1 to handle jobs on label "cloud-1"
AMI#2 to handle jobs on label "cloud-2"

If AMI#2 has idle online slaves and I start a new job tied to label cloud-1, the provisioner won't provision instances to handle the jobs on cloud-1 because excessWorkload will be negative due to idle slaves on cloud-2.