Cluster autoscaler sees pending pods but still refuses to scale

Tainted workloads remain Pending even though more nodes should solve the problem and the autoscaler deployment appears healthy.