nginx/nginx-gateway-fabric

Describe the bug NGF sets rest.Config.Timeout to 10 seconds before creating the controller-runtime manager. That client-wide timeout also applies to long-running watch requests. Each informer watch is therefore canceled and reopened every 10 seconds, even though client-go requests a longer server-side timeout. To reproduce Deploy NGF 2.6.1 with Kubernetes audit logging enabled. Inspect audit events from the NGF service account. Observe successful watch requests with both timeout=10s and a much larger timeoutSeconds value. Observe that each watched resource reconnects at the shorter interval. Expected behavior Informer watches should retain the server-side timeout selected by client-go. The shared Kubernetes REST client should not impose a shorter global request timeout. Environment NGINX Gateway Fabric: 2.6.1 Two NGF control-plane replicas Current main and 2.6.7 still assign the 10-second clusterTimeout to clusterCfg.Timeout in internal/controller/manager.go. Additional context Sho

NGINX Docs recently changed their headers to include the following fields: f5-keywords, f5-description, f5-summary, which many of our existing docs do not contain. I think these fields help with search results + AI metadata gathering for when they collect information. We should update our documents to contain these metadata fields. Additionally, many of our titles could be more descriptive. Titles such as "ListenerSet API, UDPRoute, TCPRoute, Rate Limit Policy API..." and more could be expanded upon to better tell the user what the guide is about.

gocyclo exceptions are in a few spots in the code base, most of which are just temporary satisfactions for the linter. We should refactor these functions so the logic complexity gocyclo checks for is smaller, increasing understandability and readability for developers.