Good catches, thanks. Pulling the version from each cluster’s cluster-values.yaml makes more sense than hardcoding it, and a Degraded/Missing rule covers the “synced but broken” case my alert misses. I’ll fold both into the article.

For CRDs I’d rather the build fail than skip, so no -ignore-missing-schemas. Anything missing from the datree catalog I generate myself from the CRDs and add as another schema location.