The cause of this issue was a defective transfer of DNS zone guests.core.fuga.cloud from our primary to our secondary name servers. This zone is useful for resolving host names of Fuga Cloud instances. The defective zone caused lookups of host names within this zone to fail. And because Kubernetes clusters and EMK clusters, increases the ndots threshold from 1 to 5, this zone gets appended to common host names of public services and therefore lookups also fail. Fixing the defective zone resolved the issue. We will take appropriate measures to prevent this issue from happening again in the future.