Baker's suggestion is that the only reason for wanting such less efficient means is so as to achieve the unwelcome and unwanted in the shared fiction. My OP suggests some other possible reasons too.
Great post. Suits also provides a shorter definition to explain what he means
"playing a game is the voluntary effort to overcome unnecessary obstacles."
In asking about "RPGs" I think you are asking about multiple different games, with players choosing each game in accord with or due to their adoption of some prelusory goals. Once the participants are gathered into and commence play it is reasonable to expect their prelusory goals to be in accord, although that isn't guaranteed. The griefer, for example, in World of Warcraft seems to be playing the same game while having divergent prelusory goals. The cheater is often discussed (in game studies) in terms that would make them seem to share prelusory goals with other players, while having a divergent lusory attitude (one that might be characterised as impaired.)
To ask "Why do RPGs have rules?" with that in mind seems to ask something like "Given our purposes of play why is it better justified to accept some rules over others?" Or "For what purposes of play are the rules we accept best justified?" Does that sound right? Why do RPGs each offer their particular lusory means, as embodied in their rules? The general answer might then be, because players who might feel a sense of kinship under a common banner of RPGers all have differing prelusory goals informing their preference of game texts. (I'm not saying your OP doesn't cover it: more attempting to restate it.) One potential area for further scrutiny is of course that it is principles as well as rules that form the lusory means.
The referee of a football game is obviously not a player in Suits's sense. They don't share the prelusory goal. And they do not hold themselves bound by the rules (the lusory means). They are bound by different rules, rules of professional conduct in the refereeing of a sports game. In this way, the role of being a referee - at least as far as rules are concerned - has more in common with being a police officer, a lawyer, a doctor, etc, than it does with being a player of the game.
Through this lens, we see straight away that (for instance) while a late nineteenth century Prussian free kriegspiel referee has a lot in common with a football referee, a contemporary RPG GM has little in common with them, as the contemporary GM is a player, share the prelusory goal with the other players, and has agreed to adopt the lusory means that will enable that goal (ie a certain sort of shared fiction creation) to be achieved.
The important question being raised is I believe if and how one decides to differentiate between a view of GM as referee and a view of GM as player. One lense is to entertain all of the possibilities equally that GM is a) not a player, b) a player, c) at times a player and at other times not a player, d) in ways a player and in other ways not a player.
With that in mind then, different views of GM will lead to different ideas about proper GMing. A group that view GM as referee are putting GM outside of the group of players and do not expect them to be similarly limited. A group that view GM as player go on to reason as you have. The only disagreement I have with your remarks is the implication that all "contemporary" RPG GMs are - solely - players. The game text of 5e explicitly assigns DM the function of serving as "referee" and states what that function is, which includes guaranteeing that everyone follows the rules, saying how the rules apply, and filling in gaps in rules (this last is actually left up to each group to decide how they go about it). The 5e text taken literally probably best supports d) in the list above. It's also very clear that this is not the only way and indeed many other contemporary games do not single GM out as other than a player.
To the extent that DM is referee (at the times when, or in relation to the demands of play in consideration of which, they are serving that function) they may follow principles of proper conduct which as you lay out need not include sharing the prelusory goals or accepting the lusory means. They might conduct their function in accord with principles of proper DMing much as any referee does.
In the passage I've quoted, Suits say nothing expressly about making rules up as one goes along. He was almost certainly quite familiar with Hart's discussion of the game of "scorer's discretion". So the most natural way I would expect him to approach this is to deny that rules are made up as one goes along: rather, a proper statement of the rules includes a statement of the permissions that the "rule zero" wielder enjoys. If we take "rule zero" at face value then I think the natural conclusion is that the Suits-ian description of a RPG with a rule zero is going to look much as Vincent Baker describes it
here (I've elided Baker's aesthetic judgements as best I can, which are separate from his description of the process):
with task resolution . . . whether you [the non-GM player] succeed or fail, the GM's the one who actually resolves the conflict. The dice don't, the rules don't; you're depending on the GM's mood and your relationship and all those unreliable social things the rules are supposed to even out.
Task resolution, in short, puts the GM in a position of priviledged authorship.
In short, a RPG with "rule zero" adopts as the preeminent lusory means for achieving the prelusory goal that
the GM is permitted to author the shared fiction.
Perhaps I have not put it so well, but this is the direction I have been travelling.
If GM is viewed as player (as possibly implied in the quote?), then in that case rule zero (if in play and held by this GM-player) could be like the exception made for goalkeepers in football - that they may handle the ball - and assessed for its merits on that basis. In regard to that possibility, freedom to form and modify rules during play doesn't contain any in-built restraint from disrupting Suits' construction, unless there are further rules or principles that limit the rule-forming-and-modifying to only such as continue to present unnecessary obstacles etc. I see this as similar to what Baker is saying. Remedies seen in many contemporary game texts do indeed include writing rules and principles that constrain that power (or regulate against it, including by silence.)
If on the other hand GM is viewed as referee, a word that has and continues to appear in some game texts, then rule zero can be assessed differently, i.e. in terms of how well it enables GM to succeed in the proper refereeing of a game as open-ended as TTRPG. When I read posts praising rule zero, they often seem to me to be made more from this view than the other.
