Review assigned by dnsdir.  My thoughts are more like "no objection".

I've read through the whole document but not in too much depth and - more importantly - I don't feel like an expert on this particular topic, though this draft does make sense to me.  I wanted to focus on DNS aspects here, but there are basically none: while IP(v6) addresses are very often retrieved from DNS, in this case the separation seems pretty high.

- - -
A loosely related DNS consideration did came to mind - as DNS resolvers very often have to choose which IP address to ask from a particular set (which could be spread across multiple NS names).  But in this case, I don't think that ULAs are expected to be special-cased, except maybe ignored if appearing in public DNS as nameserver addresses.  In public DNS ULAs aren't expected/recommended anywhere (as restated here in 9.2) and in case they're locally configured for some DNS subtree, I believe the resolver operator chooses by hand.  In any case, resolvers tend to have resilience by design for cases when some nameserver IPs are inaccessible or significantly slower than others.