My team has Clickhouse-Operator deployed via the Helm chart. We recently had an issue where ClickhouseInstallation CRs were not being reconciled. Upon investigating, we realized that we had misconfigured the watched namespaces like so:

configs:
  files:
    config.yaml:
      reconcile:
        host:
          queries: false
  watch:
    namespaces:
    - foo

When it fact it should be set like so:

configs:
  files:
    config.yaml:
      watch:
        namespaces: foo
      reconcile:
        host:
          queries: false

I believe that a strict schema on the values.yaml would have prevented this issue. Even if the full config.yaml is not reflected, a schema would presumably have at minimum notified us that .configs.watch is not valid and caused helm install/upgrade/lint to fail.