Answer
This points to a stray entry in the server's own hosts file that's incorrectly redirecting the snap store's public hostname to the internal cluster ingress instead of letting it resolve normally — check /etc/hosts for an entry mapping the snap store's domain to a local/ingress address, remove it, and retry the snap refresh once that redirect is gone.