As it is in Pavilion, it's really hard to have a test that works for both a large number of nodes and a small number of nodes because of how the schedule.nodes parameter is interpreted.

For jobs that leave node selection up to the scheduler, the will pick a node count based on the current cluster allocations. For testing though, we really want such a range to denote what the acceptable range of nodes is based on how many nodes are up/available, regardless of how busy the cluster is.

So, for instance:

schedule:
  nodes: 30
  min_nodes: 5

If the cluster has 30+ usable nodes, the job will be scheduled for 30 nodes. If it has from 5-30 usable nodes, it will be scheduled for however many are usable. The test will only fail to be scheduled if there are less than 5 nodes.