The game does not feel arbitrary or dictatorial from the inside. The GM decides on many things, but in truth that decision-making is bound by rule of law. Basically, the GM decides on only one thing in the game: what is beyond that next door. As long as he does not make this continuous choice arbitrarily in service of his own attempts to manipulate play, it's not even that much of a power. All the rest of the authority seemingly vested on the GM is actually under constant validation moment to moment by the rest of the players. The GM thinks that you can't swim in an iron hauberk - do the rest of the players accept this ruling - does a compromise emerge. The seeming GM authority over rules and rulings is really just presidential, not dictatorial; he's just the guy who's job it is to execute group consensus, and arbitrate disagreement. The vast majority of the time the shared creative agenda, honest will on everybody's part to respect the established fiction, and the measure of respect the GM carries are enough to ensure that the game doesn't crumble into constant bickering.
It's notable here that a major feature of modern D&D is complaints and arguing over the rules, or at least that's how the game is depicted in arm-chair theory. To me this seems like an outcome of a paradigm shift where a game that was originally fiction-oriented has come to be forced into a videogame-like objective rules framework. This means that players view the game as execution of mechanical rights and privileges: I have a right to a surprise attack under these mechanical conditions, and the even-keeled running of the game requires that this right is realized in play. In comparison, the onus of decision-making in our campaign is more on the player: does my vision of the fiction cause me to think that a surprise attack is a viable course of action here? This approach basically works as long as the players are honest and committed enough to the fiction so that sometimes they think that they deserve the surprise, and sometimes they think that they don't. It would fall apart if the players lacked that creative commitment to having challengeful play in a consistent world: we could not ask the other players to affirm and establish fictional positioning if we knew that this other guy over here, he's just always gonna say that yeah, my character's totally in position to surprise everybody. (It's part of the GM's job here to educate and encourage positive behaviors of play, of course - individual players basically stay in line with a combination of social respect and the creative purity of their own intent.)
As for teaching this to others, it hasn't been too difficult face-to-face, as an oral tradition. I've trained two other players who've been GMing with this method basically successfully. One of these has had some long-term difficulty, perhaps caused by his player base not really being on board with the described agenda over the long term. The actual method, though, is not that difficult. If anything, the difficult part is finding a group of players who genuinely believe in challenge-based gaming, a shared fiction and precedent-based, evolving rules system with a large inherent chaotic factor. Apparently many old D&D players who've gone through the 3rd and 4th edition eras are somewhat stuck emotionally on the level grind (which this version of the game obviates) and individual protagonism rights (the silent contract about the GM being there to ensure that your characters gets to "shine"). These types of players might end up playing in our campaign for quite a considerable while (dozens of sessions) before the campaign reaches a stressful point where e.g. a particularly cherished character is killed in a particularly meaningless way; at that point they might realize that our game is really just arbitrary bullshit, we don't care a whit about anybody having fun, and I didn't come here to have my character boiled and eaten by leprechauns just because of a bad dice roll. (Respectful disagreement, you understand - we haven't really had any screaming matches or anything like that.)
Regarding balance concerns, I think that we're lacking them because the campaign heavily propagandizes against the whole concept. Game balance is one of those relatively theoretical arm-chair concerns that I do admit exists as a psychological issue, but that doesn't actually do anything for the practical running of a game like D&D. That is, I do admit that people care about it, and they need to feel like their characters have a reason to exist, but that's a feature of the people, not the game. If you could just lay aside your balance concerns and focus on seeing the game from the boots of your character, you wouldn't end up noticing game balance within the game. Doing this (looking at the game from your character's viewpoint) is a very natural thing to do, as that's the very reason for the game: given that I have character X and goal Y, how do I resolve the challenge of combining this organic set of abilities with this organically grown mission? You can't waste your entire gaming time worrying about whether your character is appropriately balanced for the challenge, as that means that you'll never reach the payoff of the game. If you find yourself truly unable to stop worrying about the balance between characters or the balance between the party and the challenges, that means that you're not really on board with the fictional world we're developing: instead of accepting that this is just how wizards are in this world, and accepting that this is how many ogres you're likely to find in a single camp, you're intentionally deconstructing the activity and laying responsibility for these in-fiction things at the feet of the other players. Sure, the GM was the one who chose how many ogres, but he's not responsible for it in the sense that there would be some external measure for how many ogres there should be, and he failed it by putting in too many or too few. If you think that the GM put in an imbalanced number of ogres, instead of perhaps an unrealistic or insensible number, then you're using your own invented measuring standard that neither the GM nor the other players are.
The official party line of the campaign is that "balance" is a tactical judgement of the party of adventurers in the face of adversity. It is absolutely vital that the GM can take the same fictional problem and recognize the different challenges that it causes for characters with different means. Is the ogre guarding the bridge a mortal danger in exchange for the opportunity to pass the bridge, or is it just a potential speed-bumb, or are the characters so powerful that it's actually just a roleplaying encounter and not a real challenge at all? The players are thinking about this all the time, they need to know whether they're strong enough to take on an ogre. The GM is thinking about this as well, as the moment-to-moment chairmanning of resolution procedures requires him to be able to ask the right questions. The point is, we can say that the ogre thing is "balanced" not a priori as an arm-chair exercise, but rather afterwards: if the players found a meaningful way to relate to the event of the ogre on the bridge, and that relation involved some challenge, then we can say that it was a "balanced" encounter. The vast majority of encounters are balanced, although sometimes they are balanced towards the party running for their lives, and other times they're balanced towards the party lording it over some poor goblins.
Looking at character vs. character balance, the official party line states that Eero is happy as long as there is a random character generation method, all characters have theoretical equal opportunity over the long term, and the mechanical depiction of what individual characters are like accords with the fictional understanding. (Either we shape our understanding to match what the mechanics give us, or we change the rules.) The part about random character generation doesn't even need to be consistently applied as far as I'm concerned, as long as it's the default method. When players question it, I usually offer them variant solutions: if you feel that it would be nicer to play a heroic farmboy who was chosen by the gods to vanquish an ancient evil, we can do that. Would it be cool for you if we rolled your abilities with 10+d6 instead of the ordinary way? Or would you want to start your character from the third level? This is all consistent with the agenda in that we're choosing to play this game in this setting with these types of characters because the challenges we're interested in involve them; if a player can't enjoy the game because they don't find the challenges of playing a random average joe interesting, then of course we need to accomodate the kind of character they're interested in.
However, I am forced to conclude that the players do not actually want to play characters who are generated stronger or more experienced than their fellows - nobody's ever taken me up on the above offer. The players probably don't want unfair advantages. The closest we've come is when a player asked to play a character whose background was in high nobility; the campaign principle is that any fresh character concept is cool once, but later you'll get to do it again only on a roll of 1/6. As he was the first player who wanted to play a rich vagabond, we gave him the appropriate funds and social background. Maybe the player in question has less pride than all the players who've wanted to stick with the default chargen method, or that character vision was personally important for him, I don't know.
Regarding differences in level between characters, the official party line is that the game embraces all kinds of challengeful environments, and being the hobbit in the midsts of all the heroes of Middle-Earth is quite a valid challenge - how can you contribute with only your wits and the means of a nearly ordinary man? In practice players don't seem to mind level so much as the chaotic and very mechanically vivacious ability score system; some players now and then express some frustration at how their Fighter is less competent in fighting than the other player's differently-classed character due to large differences in natural ability. Other players and I instruct players to suck it up and do something about it; you can improve your character's chances in fights by fighting smarter, preparing better and surviving - it doesn't ultimately take that long for the well-played character to eclipse that guy whose characters always die in every session.
In practice the more usual party arrangement has been one where all the other characters are low-level, but one character in the group is considerably higher. This seems to flow more naturally from the sandbox, as high-level characters tend to have their own concerns that naturally engage lesser characters as mooks. Multiple high-level characters only congregate for a single adventure when the adventure seems particularly important. (Not merely difficult - that's a reason to send in your B-string, not a motivation to put even more valuable characters at risk!) I can't say that this would have bothered me; I get the sense that some players are sometimes frustrated by the immense mechanical difference between a 1st level and 5th level character, but this seems to be passing, and I suspect that players learn to deal with it as they learn the campaign's ideological underpinnings. That kind of frustration is ultimately in your head, nothing prevents you from playing a smart game and participating if you understand that playing the game is about ideas and leadership, not about rolling the dice. (The most frustrated players are usually ones that I would characterize as pretty mediocre in skill: they don't rely on the fiction for inventive solutions, and they have little initiative, for they expect that the heart of the game is to roll the dice where the GM dictates. Usually this sort of player gets directed to do things by players with more initiative, and this has nothing to do with the respective levels of their characters.)