I was thinking like Heroes are in Hubtown out on The Borderlands and can go hit Stonehell and Barrowmaze and maybe one other. Can always toss in like the Caves of Chaos or Palace of the Silver Princess as more land and locations are discovered.
I think this is the basic concept behind the snadbox, but for it to work I think one also needs to lay down a few piece of infrastructure.
A) Connection between overworld and underworld (dungeons).
Why are the PCs inclined to go delving in different places, how do they learn about them, what sort of information do they get about dungeon A or B that would encourage them to go there. Now obviously, as a classic sort of game, the party wants cash ... but why go to one over the other, and as always ... how to tie that cash back into the overworld. Hooks, NPCs, overworld factions with goals in the underworld and vice versa and connections between the dungeons helps. As do long term effects of ignoring dungeons - they shouldn't be static places to plunder but rather elements of the sandbox as a whole that tie into its clocks and NPC goals.
e.g. "The Dragon Hazrad the Unclean lairs in the Great Barrow of the North - its foul brood is spreading and infecting the hill people, who have begun to worship it instead of the Placid Moon ... As sessions past these former allies of Castleburg will turn to enemies, then monsters, and finally Hazrad will lead his infected army and fly covered spawn to sack Castleburg ... or so says the local raving Sun Temple fanatic... (On session 1 - by session 10 caravans are missing and villages on the edge of the hills be raided). This of course is one of several juicy rumors or claims made by various local faction members.
B) Small and mid-sized dungeons.
Generally if you are using a mega dungeon the campaign is about the mega dungeon. If you drop three mega dungeons on your map all within walking distance to a haven the party is unlikely to every delve any of them to a significant degree. 1/2 the trouble of running a mega dungeon campaign is getting the players to go deeper rather then poke around the increasingly safe level 1 (ASE has excellent ideas on how to work against this player habit btw). With multiple midsized or lair sized dungeons both the flighty and completionist tendencies of players is are (dungeons are distinct spaces, with distinct problems that have a limited number of sessions worth of content), and you can link them to the overworld more easily as well as conditioning their clocks more clearly.
e.g. "The Great Northern Barrow" is 35 rooms and depending on how far the dragon in it has advanced its schemes those rooms may be inhabited by different things. At first vermin and weak dragon spawn with a few ancient dead trying to keep them back. Then more spawn, cultists, the dragon's own dead army, reluctant hill people etc etc. The smaller size means that the problem dragon isn't ever buried at the bottom of a mega dungeon and its effects on the world can be quicker, while the party's efforts to eliminate the danger will take fewer sessions. It becomes a manageable and foreseeable immediate issue rather then so grand BBEG for when the campaign is 5 years old.
So you can't just sprinkle dungeons on a hex map - more important I think is tying everything together via rumors, faction schemes and NPC relationships.
