Kubernetes HPA stuck at min replicas despite CPU pressure

HPA reports metrics correctly (85% CPU on 3 pods) but refuses to scale past minReplicas=2. Events show 'desired replicas below minimum'. metrics-server is running, custom metrics adapter is configured. We're on K8s 1.29. The behavior started after a cluster autoscaler upgrade. Anyone seen HPA compute the wrong desired replica count? Logs show no errors, just the scale decision never fires.