This message was deleted.

FYI. I was able to resolve this. Turns out the AKS Flux configuration was constantly reconciling the default join yaml over the provisioned version. Specifically it was wiping out the env var

- name: CATTLE_FEATURES
              value: >-
                embedded-cluster-api=false,fleet=false,multi-cluster-management=false,multi-cluster-management-agent=true,provisioningv2=false,rke2=false,ui-sql-cache=false

Without this the agent goes offline