The ML team's clusters would need to override the namespaces listed in values.yaml to start fresh. In the helmfile.yaml config we have:
ml-serve-eqiad: values: - "values/values.yaml" - "values/ml-serve-values.yaml" - "values/{{ .Environment.Name }}/values.yaml"
And in ml-serve-values.yaml:
namespaces: {}
But this doesn't work since it seems doing a deep merge instead of an override, so we should find an alternative :)