What is your approach to troubleshooting a complex latency issue in a Linux-based network cluster at SciTec?