An instance group autoscaler scales on overall CPU, but latency spikes before new instances are added. Which two autoscaler settings should you change to reduce latency variance? (Select two.)

To reduce latency spikes before new instances are added, you need the autoscaler to react more quickly and provision instances sooner. Decreasing the cool-down period allows the autoscaler to consider scaling actions again sooner after a previous scale-out event, preventing delays in responding to continued load increases. Decreasing the autoscaler's target CPU utilization means the autoscaler will trigger a scale-out event at a lower CPU threshold, provisioning new instances proactively before the existing instances become overloaded and cause significant latency. Increasing the target CPU utilization would have the opposite effect, delaying scaling. Adding a label or removing health checks are unrelated to autoscaling responsiveness or thresholds.