I've also got a weird state where 2 out of some 10 nodes are having issues, but others are running fine. It started going after upgrading from (k8s) 1.21 -> 1.22.