I think the monitoring chart might have a bug in the prometheus-adapter config rules. Currently, if you tail the logs, there is always some amount of timing out. Looking upstream to the prometheus-apdater repo, there are a few different issues about this, but no clear solution other than some vague hints that this config might be overloading the way the chart is structured. I'm going try to dig in more, but would appreciate anyone that is able to verify by checking your prometheus-adapter pod logs for

timeout or abort while handling: method=GET URI="/apis/custom.metrics.k8s.io/v1beta1"