This post ties in with yesterday's Traveller post, but also with something I'm working on for Necrocarcerus 1.3.
Sandbox settings typically have two different scopes of play during a session, and moving between them well is critical to pulling a sandbox. I'm going to call these two scopes the "strategic" and the "tactical" scopes for clarity.
The strategic scope is the scope where the player characters consider the list of options for missions, tasks and goals. In play, this scope often involves consulting maps and lists, but it doesn't exclude roleplaying. PCs may consult their patrons, discuss their options in character, check in with sources and contacts, etc. This often even includes the actual travelling portion of game-play. The important thing is that they are not necessarily committed to any particular course of action. Deciding whether to smuggle xenopornography to the dolphinoid revanchist militias of an interdicted waterworld (a real example from a Traveller game I ran) is the kind of decision you make at this scope.
I occasionally see this section written off as "prep" with the suggestion that it should be elided or compressed, but I think this does a disservice to the possibilities of play it generates. A common referee mistake here is to undersupply the PCs with information they need to evaluate or predict possible consequences of their actions. While one doesn't need to simply hand them everything they want to know without effort or cost, it's useful to explicitly ask them the critical factors they need to decide between courses of action, and then detail how they can obtain this information.
The strategic scope helps fix the details that feed into the tactical scope. It determines time pressures, goals, resources, allies and enemies - basically it generates the framework of the individual adventures.
The tactical scope really begins when the PCs make a decision that can't be undone without abandoning the goal. So, when they dock at the space station to investigate the SOS signal, or when they make the first move to steal the nuke snuffer from their rival, or whatever. Here we enter the traditional scope of play - usually involving a specific location or small set of locations, where the PCs describe their actions individually and shoot their lasers at enemies, etc. The tactical scope is where adventures happen.
In Traveller, much of the procedural generation material exists to support the strategic scope of play, rather than the tactical scope. I believe this is true of most sandbox games I'm familiar with. This doesn't mean the tactical scope is unimportant, but strong support for the strategic scope is a feature that we use to declare a game is a "sandbox" instead of some other type of play structure.
The part I want to back up to for a moment is that transition between the two, since I think this is the part that trips people up the most. It involves shifting gears between two different styles of play. Some examples of games that have both of these scopes in them and clearly demarcate them are Burning Empires, and Stars Without Number (especially the Darkness Visible supplement). I recommend checking either game out for more information, but I'm going to just mention them here rather than go into great detail about either one.
The demarcation point between strategic and tactical scopes of play is the choice that cannot be undone. This line of demarcation is taken from dramatic writing (a choice that cannot be undone is the transition point between acts of the story in film and plays, specifically). "Undone" doesn't mean the PCs can't leave the dungeon, fly away from the asteroid, whatever, but that they can't do so without some cost or risk that would not occur had they not engaged with it in the first place.
You can move as freely as one likes between the two scopes in actual play, so long as this line of distinction is maintained. If it isn't, you'll find people start getting confused about their options. It helps to call this out a bit in play, often by citing the obstacles to disengagement before the PCs fully commit. "Once you dock with the space station, it'll take a half-hour to disengage and break contact if there are any problems" or "If you go and talk to Murderous Marco and he offers you a job, you'll either have to take it or there will be trouble, regardless of how bad the terms he offers are."
This also helps in defining the scope of adventures. Knowing that the adventure proper must begin with a choice that cannot be undone, you can design your adventures to clearly begin with them, instead of just kind of drifting into suddenly having an adventure, which is a common mistake referees make as they try to manage the two scopes.