Well, IPs of these instances do not match with any pod IP or worker node IP. Hence we can rule out it being a cni or service mesh issue, right?