I'm digging through some of the upgrade logs from that Longhorn install. I'm wondering if what happened was that we installed Longhorn via CLI + Helm Chart, but then I went to upgrade through the Rancher UI. I'm saying this because 1) in another cluster, we have an install of Longhorn done via UI that includes an install for the longhorn-crds and longhorn itself 2) when I did the 1.4.0 to 1.4.4 upgrade via the Rancher UI, the logs helm logs show helm deleting a disturbing number of CRDs.