I have been thinking lately that there is a contradiction, a tension, in the way that we play ACKS (and probably in other OSR games as well). The two competing modes of play are the open-table game and the domain game. The demands on the structure of the game that each imposes are inimical to the other.
The open table more-or-less requires episodic play. It's possible to run a coherent non-episodic game with an unreliable cast, but it's a lot of work. Beyond the Black Gate suggests a number of approaches to episodic play - relic retrieval, monster hunting, and the megadungeon. Justin Alexander cites the megadungeon as his path back to open-table gaming. The Western Marches, long the canonical 3.x open-table game, actually has a lot in common with the megadungeon. It has the same difficulty gradients, hidden treasure rooms, high lethality, unclearability, and internal logic one would expect of a megadungeon.
These styles of play (except for the relic retrieval) are well-designed for denying players their impersonal resources. In both the megadungeon and the Marches, mercenaries are inviable due to morale concerns (it's considered suicidal to go there and you can't hire), and establishing anything resembling a domain is impossible due to the unclearability / restocking property.
This is not accidental. Episodic play starts to get really weird when you add domains and mercs, mainly due to time, continuity, and boundary conditions. If you're not at a session and they elapse a month of travel time, do you need to pay your mercs their monthly wage? Does your domain generate income? If the next party is totally separate, maybe not, but what if someone who was at the last session (and acquired XP and treasure, and whose 'timeline' was advanced that month) is at the next session with you? If another party accidentally leads a goblin army back to your domain during a session where you're not present, what happens? In strict episodic play, it's very easy to handwave a character's absence for a month, but in the domain game, time and continuity are both sort of important.
My conclusion from this, then, is that we have another subtle shift in play in B/X and its derivatives. Not just within the rules, but within the structure of the party and the game. At low levels, when you're dungeoneering, open table works fine. You can swap out players and characters and it's all good. Expeditions to the dungeon and back are only a day anyway. When you start doing wilderness adventuring, continuity and time become more important, because you start accumulating mercenaries and expeditions get longer. Switching to the relic retrieval and monster-hunting episodic modes mentioned above can keep a wilderness-level game open for a while, but disparities in level and personal power will begin to emerge between reliable and unreliable players, and continuity and time headaches will start to accumulate if mercenaries are in use. It might be sensible to fork the game at this point, with reliable wilderness players who are accumulating mercenaries splitting off to form their own game with a separate, coherent continuity from the continuing open table game. It seems like such a "forking" measure would be nearly necessary by domain level.
ACKS, for its part, does not make running an open table easy even at low levels. The primary offenders here are the market rules and henchman wages, which require timekeeping from 1st level (though I suspect that many a henchman's monthly wages have gone unpaid and unnoticed in my games in the past). The spellcaster repertoire-swap and spell-learning times also encourage timekeeping.
At the end of the day, I might just be disillusioned with both open table's continuity and timekeeping issues and the domain game's paperwork, though. I keep trying to run open tables at around the wilderness level and they keep being pains in the ass. I'm sorely tempted to try falling back to a small, reliable party of 2-4 players for a wilderness-level game.