Two blogs that I highly respect, Grognardia and Greyhawk Grognard have very interesting posts on Megadungeons here and here.
The basic problem is that we never had a published version of any classic Megadungeons we heard about Blackmoor, Greyhawk, El Raja Key in the original rules they used. We had various later version but they all suffered from the "We made shit up for this release" problem.
Much like the beautiful Darlene Map was made for the Greyhawk Folio and doesn't not represent the map Gygax actually used for his campaign. Although it was said later he adopted it as his own.
In the last couple of years, because of the Internet, a lot of people wound up comparing notes with help of commentary from the folks that ran stuff back then (Gygax, Arneson, etc). The general consensus is that the problem is the original dungeon don't exist in the format that we think they exist in. The original dungeons are mostly a map with notes and few detailed rooms. Often coupled with some method of random generating room contents and inhabitants. It explains all the examples we currently have (Blackmoor from FFC, Tegal, etc)
However what dominated the publishing market was not those early example but the fully fleshed out Tournament Dungeon like the G series, Tomb of Horrors, the Ghost Tower, the A series, etc. This became THE format in which to present an adventure module.
Suddenly those old notes looked unpublishable much to the detriment of our hobby.
While both James and Joesph make good points they both miss the mark.
See the greatest enemy of a DM is time. Time to prepare for a upcoming session. It is only natural that the early (and current) DM try to only write down the minimum needed to run the session or campaign.
This press for time create a market for publishers to provide GMs with products that save time. Adventure module, aides, settings, etc. The first thought of a publisher is "Hey! I can spend the time fleshing out everything for the GM. Everybody will find that valuable." It logical and easily understandable. But it is also probably the worse thing one can do for these old dungeons.
One of the initial appeals of the RPGs, as Greyhawk Grognard points out, is their dynamic nature. Now you need some base level of detail but beyond that you need to make shit up in response to what the player are doing.
I think the notes found with the few examples of the old Megadungeon are sufficient with one important caveat. The GM needs to be taught how to use them. And that what lacking from the old days.
The way to tackle this for a mega dungeon not just be a map and a collection of sparse notes. But also have chapters on teaching somebody that has never run this type of dungeon before how to run it. They need to be written not as a series of detailed room description but how make up detailed room descriptions ON THE FLY. How to keep thing consistent (if that a goal) and all the other issues.
Then suddenly the sparse notes format of the old megadungeons become publishable in a affordable format. New Megadungeons can be made within a reasonable time. They become a springboard for the GM and player's creativity rather than the straight jacket of the tournament style dungeon.