Bad scheduling of jupyter pods

The jupyter user pods should be scheduled with round robin principle. Currently all pods are scheduled on one server until it runs out of capacity. Only then pods are scheduled on another one. Here is a thread with the same issue.