Alright I need some help and maybe just to vent a bit. I've been trying to configure split tunneling on Linux for a data-scraping project and the results are just not matching the logic. I'm running a WireGuard setup on a VPS, and my goal is to route only the scraping script's traffic thru the VPN, keeping everything else on my local connection. The problem is latency. When I use the app-based split tunneling in my paid VPN client it's fine, but when I try to do it manually with routing tables and policy-based routing for more granular control on my Linux box, everything either goes through the tunnel or nothing does. I've spent probably 8 hours this week checking iptables, wg-quick configs, and using netstat to trace routes. The data is inconsistent - sometimes the script's outbound IP is my home IP, sometimes it's the VPS, with no changes to the config. It's driving me nuts. It all comes down to the human connection, but right now the connection I'm having is with a terminal window and it's losing. Has anyone actually gotten manual split tunneling to work reliably on a headless Ubuntu server for a specific application? Not just browser splitting, but for a Python script or a Docker container. Looking for specific CLI tools or config snippets that gave you solid, testable results. My current setup is failing the most basic test: routing a curl command through wg0 while pinging 8.8.8.8 locally. One of them always breaks.