Recently upgraded from kubernetes plugin 1.14.3 -> 1.14.8, so not sure which version the change was introduced.

 Tested and found that this behavior was introduced with 1.14.6

Seeing an overeagerness with node provisioning with k8s nodes. Happens most often when there are already existing k8s nodes jobs running

to repro

1. Create 3-4 jobs running with a k8s node label. I just used sleep 600 in my build
2. start one of the jobs
3. while that build is running, start the other jobs
4. observe that more nodes will spin up than necessary and so some will sit idle before being terminated
5. observe in /var/logs/jenkins/jenkins.log messages like this, negative workload: 

   // INFO: Started provisioning Kubernetes Pod Template from kubernetes with 1 executors. Remaining excess workload: -0.183