I treat the network as part of how an application works, not just an infrastructure layer. I follow traffic from DNS and TLS through the reverse proxy and load balancer to the target service, which makes it possible to diagnose problems instead of restarting one component after another.
The same goes for traffic between services inside a cluster: a service mesh (Linkerd, Istio in ambient mode) gives every connection an identity and mTLS, and routing, canary, retries and timeouts are described with one standard, Gateway API, at the cluster edge and inside it alike.
TCP/IP, DNS, HTTP, HTTPS, TLS, Routing, NAT, Firewalls, VPN, Reverse Proxy, Load Balancing, Nginx, Traefik, Gateway API, Envoy, Service Mesh, mTLS, Linkerd, Istio, Network Debugging