Based on discussion on (closed) #207, this issue is a placeholder for figuring out if the nodePort (felt as an escape hatch) can be obliterated with a dns setup that gives the container runtime access to in-cluster service dns resolution.

In scenarios where cni network policies are in place, those policies will need to be considered, too.

Loose ideas:

• modify host's resolv.conf (delegate resolution of well known dns name to CoreDNS)