When deploying a self-hosted Temporal Server on Kubernetes, request latency can be an important performance indicator. To achieve reliably low request latencies, it's essential to monitor and optimize CPU usage, as excessive throttling or limits can lead to increased latency. One effective strategy is to set the GOMAXPROCS environment variable to match the number of available cores, allowing Go to make better decisions about concurrent processing. Additionally, disabling CPU limits altogether and using dedicated nodes with taints and tolerations can help reduce latency spikes during deployment rollouts. By following these tips, users can improve the performance and reliability of their Temporal Cluster while saving resources.