6. The test code remains difficult to review because I can't see the forest for the trees due to the dozens of S->S1 node name changes. These name changes are unrelated to the new feature so please separate them into a different prerequisite patch so we can just focus on what changes were made just for --drop-all-publications. I know you already said you are working on it [1-#7], but multiple patch versions have been posted since you said that.
I don't see the point of renaming S to S1.
I also don't get re-defining the existing S tests to include --drop-all-publications and then go and add a new test that excludes --drop-all-publications.
Just name the new subscriber D (or pick some other random letter, it isn't like we are encouraging meaningful variable names here) and have it test the behavior of --drop-all-publications. Given the existing design, the introduction of two new user publications on P initially shouldn't impact existing tests (and if it does, seeing that change in the existing tests would be a good thing). An extra check or two against S should suffice to prove non-deletion.
In short, this probably needn't touch any existing tests, just the shared setup.
Also, can we explain why it is important to use --verbose --verbose? Obviously it shows more information, that what repeating --verbose is defined to do. (Sure, that isn't new to this patch.) But why is it important that this specific command use it when none of the others do?
