Would you say it's closer to #2? The details you just described seem more like they might fit that category. I was pulling out #1 from the impression I got about motive and approach.
Here are your (2) and (3):
(2) Alternate Universe Setting: You start with all of the assumptions of the published world (according to whatever materials you have available to you) and then selectively make massive changes to a variety of things, which might include well-established world assumptions, maps, races, NPCs, past timeline, maybe even playing with a different ruleset.
(3) Canon with Selective Changes: You start with all of the assumptions of the published world (according to whatever materials you have available to you), and assume anything that doesn't come up during your campaign adheres to those assumptions unless otherwise stated. You make selective changes to parts of the world you feel should be changed to better fit your personal vision, including past timeline, NPCS, and other relatively minor elements.
Both talk about
starting with all of the assumptions of the published world. And then
making selective changes. The difference seems to be the degree of change ("massive changes to a variety of things" vs "minor elements").
If I was creating your scale, I wouldn't express the rankings in the same way.
For instance, you say "even playing with a different ruleset", as if this is the most massive of massive changes. Whereas for me it is relatively trivial. I've run GH games using AD&D (with various houserules); a homebrew AD&D variant; Rolemaster; and Burning Wheel. The different systems of course produce different play experiences - that's what different systems are for - but I don't feel this is a massive change in the
setting. It is not very important to the identity of most GH NPCs, for instance, that they be particular D&D classes (as opposed to fit certain fantasy archeypes that all these games support); and GH, like most fantasy worlds, is full of events and phenomena that don't seem to be particularly derived from an application of the D&D ruleset. (Eg the Invoked Devastation and Rain of Colourless Fire.)
I've likewise run OA games using both AD&D and RM. The latter game integrated, among other things, bits of a free Palladium scenario I found online back in the late 90s, and bits of a Bushido scenario I picked up second-hand. This is another case where system is secondary, and what matters is being able to mechanically express the relevant tropes and themes.
In my D&D GH game orcs and hobgoblins were different (because the D&D MM told me so). In my RM GH game (and to the extent that the issue might even arise, in my BW game) orcs and hobgoblins are the same people (so the differences between the hordes of Iuz and the hordes of the Horned Society are cultural and political/social, not "racial"). This doesn't strike me as a very big deal either. I've also treated hill giants and ogres as the same people in my RM game - another change that, in my view, makes little difference to the fiction of the setting. (The difference between hill giants and ogres in D&D is mostly about HD - a mechanical thing - rather than theme, trope or other fictional element.)
But change the history of the Suel Imperium and the Great Kingdom, and
then I'm not sure you have GH at all (as opposed to a different setting using the GH maps). Because these are what is distinctive about GH - it's version of the classic pulpy ancient empire (Acheron, Stygia etc in REH) and the Hyborian kingdoms (Aquilonia, Ophir, etc in REH). So what you seem to list as a minor element to me seems pretty central to the setting.
My other departure from your approach to ranking pertains to "assumptions" and "changes". The way you present it seems to suggest something like an editing process. Whereas that is not generally how I do things. I don't "assume" and then "change". I use certain stuff (eg maps, a description of some place or person, etc) and that becomes part of the setting. And whether something else written in the setting book, or some other supplement I own, or in some future supplement that I purchase, is part of the campaign will depend on whether or not it comes into play and seems worthwhile.
For instance, I'm sure that many many OA games have been played without having any occurence of the Animal Courts described in OA7 Test of the Samurai. But in one of my OA games these were quite important to the game, in part because of the way I was integrating OA7 and its backstory into the bigger picture of the campaign (including connecting it to the independent-as-published OA3), and partly because it turned out that one of the PCs (a fox able to take human form) whom, at the start of the game, we had assumed was a fox aspiring to humanity, was in fact an animal lord who had been banished from Heaven and cast down to earth in the form of an animal and with his memories therefore mostly lost.
In the same campaign the Dragon Claws from OA5 figured, but not Mad Monkey. (And we never considered the Cat Lord either, despite the fact that the original OA book indicates that he is part of the OA cosmology.)
So for me it's a process of "filling in" details, either in the course of play or as prep for play, and drawing this from whatever seems interesting and worthwhile, which might include the setting book or some other book. A lot of the details in setting books aren't that important for any particular game, and so it is of no real consequence to ignore or change them; and, conversely, some details can turn out to be important in one game though they would mean nothing to most users of the setting. For instance, to refer back to OA7, I'm sure many OA gamers have not thought much about the Peachling Girl and her riding tigers; whereas in my OA game the Peachling Girl turned out to be a significant NPC, and Momoben Forest tigers also turned out to be some of the more important animals of the setting. (Because there was a PC who liked to summon them using his druidic magic.)
But if a game had
nothing like Animal Courts, or foxes who can turn into humans, or Peachling Girls, or riding tigers, or Dragon Claw, or Mad Monkey, then I'm not sure that it would be an OA game. Because these are the tropes that make that sort of game what it is. (Unless you're going for a very "hard"/"sparse" historical feel - but then you have to have samurai and daimyo, or emperors and eunuchs, or some other tropes that mark it out as an OA game. A game in which everyone has OA classes and OA gear, but the actual action is indistinguishable from The Sunless Citadel, wouldn't seem to me to be much of an OA game, even though none of the setting backstory has been tampered with.)
Another tangential concept that has come up in this thread and others is how personal attitude towards settings can influence how you feel about using them. The same person might abhor deviating from canon in Star Wars or Middle-Earth, but see the Forgotten Realms as a big book of suggestions for making their own D&D world. So, for some elements of the discussion, we can probably add the consideration that a lot of it comes down to how you feel about the setting. If you really like a setting's implementation you are probably less likely to change it
I've never run a game in Middle Earth, and probably never will.
Middle Earth, to me at least, is primarily a story-telling vehicle, and the stories have already been told.
The Hyborian Age, on the other hand, could make for an excelleng FRPG setting, I think - it's not so much a vehicle as a backdrop for whatever pulpy stuff one wants. But then Greyhawk gives me much the same, without the Conan baggage, so I use it instead.
But if I was to run a game set in the Hyborian Age, I certainly wouldn't be worrying about what NPCs I introduced, or what gods and cosmology. REH just made up the stuff he needed to make his stories work - as a GM, why wouldn't I do the same?!