We’ve had scaling lag issues during traffic spikes. Nodes taking too long to boot whenever we need to scale. I tried using hibernated nodes, but Karpenter takes about the same amount of time to wake them up. Then I realized my bottleneck is the image pull, I tried fixing it with an image registry, which sometimes helped, but other times startup time was exactly the same. I feel a little stuck. Curious what others are doing to keep autoscaling responsive without wasting resources.