A cluster autoscaler adds nodes and pending pods still stay unscheduled

New nodes appear, yet the scheduler and autoscaler disagree about the effective label set available for the target workload.