Karpenter vs cluster-autoscaler for EKS spot fleets — real-world cost delta?

We migrated from cluster-autoscaler to Karpenter on our EKS workloads last quarter. Spot interruption handling is noticeably better, but we're seeing ~12% higher on-demand spill during peak hours because Karpenter's consolidation logic is more aggressive about bin-packing. How are others tuning the provisioner profiles to balance cost vs availability? Specifically: what consolidation TTL and drift detection intervals are you running in production?