There is some discussion among the OSR blogs about sandbox campaigns. Long-story-short, I agree with Chicago Wiz completely. (I'd tell him so on his blog, but the DayJob filters are blocking the comment entry script. If I have time to tonight I'll cross-post this there.) Additionally, having read his list of quests I realized my Southern Reaches campaign has room to grow - there are layers I need to add.
As to the difficulty in setting up or running a sandbox campaign, I find they take about the same amount of time to prep/work, just spread out differently. It definitely does require the DM to think on his feet more and faster and if you as a DM cannot do that or feel uncomfortable doing so, this is not the campaign style for you.
In the comments, JB states he has problems dealing with "reconciling MULTIPLE players wanting to do DIFFERENT things at once". This is a real issue, but not one inherent to sandbox campaign. This is a player group issue and would be an issue in any style of campaign. In story-driven campaigns it can be muted as there is a clear path for the party to follow, but even then the individual players may want to pursue things in different ways. This is herding cats at its finest.
I run a West Marches-style campaign and it does require the PLAYERS to agree on what the group is doing on any particular evening. Usually, the discussion is based on what they want to achieve that night (loot, xp, or exploration/knowledge) and who is available ("we aren't going to the trap-heavy Terrace if the rogue is not available"), but ultimately, the players have to decide to agree on a location or nothing happens.
A sandbox DM has to be willing to let nothing happen if the players don't get their act together and cooperate.
If the players can't work something out (we'll go here this time, but next time we are going where I want), then everyone's time is wasted. Sometimes the DM may have to step in and say "look, 5 out of 6 of you want to go to X, so I'm running an expedition to X tonight. Player 6, please accept that or we'll see you next time." If Player 6 is a repeat offender on this, you could run a separate game on a different night or just have The Talk with that player and explain that this is not the group for them.
Now if all the players want to go different directions and none will compromise, then you've got a pile of jerks and there's not much you can do with that. Other than run each one separately against a big monster(s) - after they have all died alone (possibly multiple times), they should start seeing the benefits of working as a group, not a bunch of individuals.
Sandbox campaigns should not be a surprise for the players. If the DM has explained what style of campaign is being run (the players have to go looking for adventure) and stayed at the table anyways, they have implicitly agreed to go with the precepts of the campaign. One of these precepts is that the party will agree on an objective for the session. In a pure West Marches-style game, the PLAYERS have to arrange the game times and the DM just shows up to run. The way most people seem to run it, the DM still arranges the time and place. This leaves the players with the job of figuring out where they want to go as a group. And if they want to go off the grid, it is up to them to warn the DM before the session so the DM can have something ready. If they spring it at the last moment, they may have to accept a "Under Construction" sign until the DM gets that location ready.
Now to follow the Joesky Rule, here is a random table:
1 Texas Drought: 1 in 6 chance of a wildfire happening.
1-2 Started by a campfire
3-4 Started by lightning
5-6 Started by an idiot with fireworks
2 Mississippi Rain: 2 in 6 chance of More Rain
If More Rain:
1-2 1 in 6 chance of flooding, +1 per week of More Rain
3-4 1 in 6 chance of tornado
5-6 1 in 6 chance of river barge hitting a bridge (1-2) or other river traffic (3-6)
3 Ohio Late Snow Storm: It snows for 1d6 days, more than a month after winter is "over"
1 It sticks to the ground and piles up
2 It sticks to the ground, but melts the day after the snow stops.
3 It melts immediately and may cause flooding (1 in 6)
4-6 It melts immediately, but no one comes in to work.
1 Arkansas: low intensity, multiple quakes. Caused by drilling/dwarven mining (based on campaign).
2-5 California: Wakes people up, but then they go about their business.
6 Alaska: 1 in 6 chance of volcano forming