So the options were: fix your applications so they don't allocate too much, don't run those applications on k8s or put thousands of valuable engineering hours into adding an incredibly complex failure mode to Kubernetes. They chose... poorly.
#kubecon