A common lament amongst old schoolers is that Gary Gygax never published a "complete" version of Castle Greyhawk. You also occasionally hear similar laments about Dave Arneson's Castle Blackmoor or other iconic megadungeons from the early days of the hobby. I used to make such laments myself, so I have a lot of sympathy for those who wish they could see the maps and contents of these foundational adventure sites.
As I've studied the history of the hobby, though, I've started to understand exactly why we've never seen a "proper" megadungeon published to date: it can't be done, at least not easily. Now, before everyone jumps all over me for saying this, let me clarify a few points. When I talk about a "proper" megadungeon, I'm not just talking about a big dungeon. Now that the old school is the flavor of the month, lots of people use the term "megadungeon" without understanding that it's more than just a module on steroids -- a "super-module" to use TSR's late, unlamented phrase. When I say "megadungeon," I'm talking about what Trent Foster calls a "campaign dungeon," which is to say, an adventuring site so large, open-ended, and dynamic that it becomes a campaign setting unto itself -- a kind of "dungeon wilderness," if that makes sense.
When most people think of a "dungeon," they expect a set of maps with a key describing rooms and their contents. A megadungeon, by its very nature, can't be detailed in the same way. It's a lot more "impressionistic" and relies heavily on ad hoc adjudication by the referee, as the players explore it. Not all of the megadungeon's rooms are inhabited at any given time -- this is important -- and many of their inhabitants might change, depending on player action, referee whim, or the luck of random rolls. Likewise, even the geography of the megadungeon might change, as the referee adds new sections, closes off old ones, or otherwise alters what the characters have experienced to date.
All of these factors make it hard to present a megadungeon in a form that's easily usable by anyone other than the referee who created it. My own Dwimmermount megadungeon, for example, consists of a collections of piecemeal maps, quick notes, and random tables. If someone asked me to show them Dwimmermount, they'd get a stack of papers that'd probably make no sense to them, because their "organization" isn't according to the scheme of a published module but rather according to the needs of the sessions I'm playing with my players. As the campaign wears on, certain aspects of the megadungeon become set in stone -- like the "cleric tree" that bears fruit that act like potions of healing -- but many other aspects change from session to session, often without the players realizing it. These changes are sometimes purely whimsical on my part and other times they're based on a growing sense of what would work better as a challenge.
In every case, the changes are in response to play and it's this quality of megadungeons that makes them hard to put into a published form. As campaign settings in their own right, the best one can hope for is a "snapshot" of the megadungeon at any given period of time. Megadungeons aren't meant to be "cleared out;" they are constantly growing, changing, and re-populating. They're not just a bunch of maps with keyed rooms and to treat them as such is to miss the whole point of them. Rather, they're what Joseph the Greyhawk Grognard calls a "tent pole," a hub around which an entire campaign evolves. That's certainly been the case with Dwimmermount, as almost all the action outside of the megadungeon itself has been in relation to it. Dwimmermount is thus both a physical location for adventuring but also a "cultural" one, which is to say, a background against which other events and interactions take place.
If you read any descriptions Gary Gygax ever gave of the evolution of Castle Greyhawk, you'll quickly see that the Lake Geneva megadungeon was very similar to what I'm describing. Gygax kept notes about the dungeon, but there was nothing at all equivalent to a module write-up for its myriad levels and sub-levels. The same could be said for Arneson's Castle Blackmoor -- and we can see this in First Fantasy Campaign -- or in the underworld beneath Jakálla in Barker's EPT campaign. These megadungeons only acquired certain of their features through play and, even then, those features could and sometimes did change, which is why Philotomy's notion of the "dungeon as mythic underworld" rings so true even to a Gygaxian naturalist such as myself.
A megdungeon product built around the way these dungeons were actually used would be akin to many boxed campaign settings, where you get "high level" information about countries and rulers, important local sights, and places of interest, but leave the specific details up to the referee as his players encounter them. Now, I think such a product would be very interesting, but I fear that many gamers, raised on a steady diet of "complete" adventure modules, would find this approach unsatisfying and confusing. And their feelings are justified in as much as we've never really seen many examples of proper megadungeons published in the last 30+ years. Paul Jaquays's Caverns of Thracia probably comes closest and I am told that at least the first Undermountain boxed set takes a similar approach, but I cannot say that with any certainty.
All of the foregoing then is an explanation of why I no longer fret much about never having seen Gary's Castle Greyhawk in printed form. I simply don't think such a thing would ever have been possible and any attempt to present a "Castle Greyhawk" trapped in amber would necessarily feel inadequate. That's the nature of the beast and therefore I think the only way to experience a proper megadungeon is to build it yourself.