Hey team, we upgraded RKE2 from v1.26.7+rke2r1 to v1.27.16+rke2r2 which also upgraded Calico from v3.26.1 to v3.27.3 and tigera-operator from v1.30.4 to v1.32.7. Since the upgrade, tigera-operator has been in a permanent reconciliation storm (~150 loops/min) and periodically all 30 calico-node pods simultaneously lose their Typha connection causing a cluster-wide network outage. Felix takes ~16 minutes to resync after reconnecting, breaking kubelet-to-pod path on all nodes.

maxUnavailable: 1

is set but not respected — all pods disconnect at the exact same millisecond across all 3 Typha pods. Filed RKE2 bug: https://github.com/rancher/rke2/issues/10068 Has anyone seen this behavior after upgrading to v1.27.16+rke2r2? Any known fix or workaround?