Question is: is the actual code so useless that it can't even be a 1.0 release?
What's committed suffers from a design problem and cannot work correctly, nor can it be fixed without an API change and significant revision. The revised version I posted yesterday is that fix, but it's new code just before beta1. It's pretty much equivalent to reverting the original patch and then adding a new, corrected implementation. If considered as a new feature it'd never be accepted at this stage of the release.
A lot of (complex) features were introduced with the notion that will be improved in the next version.
Absolutely, and this is what Petr and I (and Andres, though less actively lately) have both been trying to do in terms of building on logical decoding to add logical replication support. This is one small piece of that work.
It's a pity since I'll have to maintain a patchset for 9.6 for people who need physical HA support for their logical decoding and replication clients. But it doesn't change the WAL format or catalogs, so it's pretty painless to swap into existing installations. It could be worse.
However, if the code is buggy or there are serious API problems, revert them.