Ken Harrison has been posting about megadungeons recently, and asked about the experience of others playing in/GMing megadungeons.
As a GM, and reflecting what I see of my player's approach, is that megadungeon play is a mix of the immediate, and the cumulative.
The immediate in play is what you're doing right now, and what affects you right now.
- your supply of resources
- your immediate goals
- your take this expedition (loot, knowledge, experience, whatever - your gains)
- the obstacles and combats you must face
all of this is right now. But the cool part of a megadungeon is, right now can feed into later.
The cumulative in play is what you've managed to build up in expedition after expedition.
- your maps
- accumulated knowledge of the dungeon
- cleaned out safe areas
- attrition of monsters
- demolished doors
- newly discovered entrances and exits
- set traps
What I like about this is that every trip is both immediate and cumulative. Expeditions have their own problems right now, and their goals right now, but it matters what you accomplish. Even a bad trip, with casualties or a lack of treasure or minimal exploration, accomplishes something for the next trip. It might add a new room to your map, reveal a new infestation of monsters, pick off a few critters from a larger group.
Conversely, even a successful expedition can give you some bad long term results. Perhaps you weaken one group enough that they get displaced by foes more dangerous to you. Or you pick off a few of their guards but they get on to your M.O. and capabilities.
A fight with a monster you can't beat can result in a retreat and a return with more information or firepower. A mission can be planned around things found in one trip that can't be exploited right away, but deal with later.
You get this mix of the tactical and strategic, too. It really matters how you take down doors, guard your six, move in the dungeon, and so on. But it also matters what you plan to do and how you arrange your overall trip, how you make this trip feed into the next trip ("We clear these guys, and next time no one hassles us on the way to and from level 3"), and so on. Killing a chokepoint monster might make a new area accessible. It might make an area safe to explore because the threat of death on the way out is now gone.
So your choices, tactically and strategically, matter both immediately and cumulatively.
This means you can make a tactical choice to back off from a fight, and pay for it later as the threat still exists. Or you could be making a strategic dent in the opposition by whittling away their resources by attrition. Or both - you can be cutting them down, reducing their numbers, but giving away knowledge of your capabilities and relative threat level.
All of this is because it's a reactive environment, not a static one, and it's not a one-trip dungeon but a location you return to many times. This works for the GM and the players, who can mix bold immediate goals with cautious long-term ones. Or the reverse - plan a series of cautious immediate goals, which lead cumulatively to a long-term benefit from the dungeon.
That's what I really like about running a megadungeon. I never did it back in the day - I came of age in the Module Era, and we ran (and even made our own) modules. But yeah, every session matters both now and later, and that's a real plus to megadungeon play.