Operator crashes when watching non-existing namespaces

[rlapond]

Bug Report

What did you do?
Deployed ECK via Helm and only specific namespaces are managed defined by the managedNamespaces value in the values.yml for the Helm chart.

After deleting a cluster and the namespace, but the namespace is still listed as managed in the Helm values. The operator crashes with the errors: "Can't list resource in namespace" and "Failed to sync cache" and the pod gets restarted.

What did you expect to see?
The Operator stays healthy, but only generates an error "Can't list resources in namespace"

What did you see instead? Under which circumstances?
After the namespace is deleted. The Operator is happy for x amount of time and in our case after 5/6 days the Operator keeps crashing every 2-3 minutes.

Environment

• ECK version:

  2.9.0

• Kubernetes information:

  On Premise - Rancher

$ kubectl version
 v1.26.12+rke2r1

• Logs:
  I didn't save the exact logs, but these 2 come to mind:

Can't list <resource> in <namespace>
Failed to sync cache