Hey Mike,
A few thoughts, sadly mostly about process rather than merits. sorry about that.
1] You've got HttpBis at the top of that document - but I don't think this is something httpbis would adopt (the word http does not appear anywhere else in the doc other than url references) as its scope is really about DNS libraries servers and registrars, not about the http protocol.
2] Given that this is about DNS libraries servers and registrars, don't you think this should be part of the DNSOP wg discussion?
3] This seems to boil down to NEWRFC changes a bunch of RFC6761 SHOULD's to MUST's. Obviously publishing NEWRFC doesn't make anyone non-compliant with 6761, so how does this help in making you confident enough in localhost==127.0.0.1 to use that in secure contexts? i.e. what's the indication the name is resolved according to NEWRFC rules?
4] you say '"localhost names will always resolve to the respective IP loopback address". That suggestion, unfortunately, doesn't match reality.' but you don't provide any background. That reality will be intentionally prohibited by this document so I would like to understand the instances where this happens a little better, what they are accomplishing, etc... you've obviously got some examples. Thanks.