I hear it expressed fairly often that the principle reasons for a sandbox campaign being difficult to run is the vast number of possibilities that present themselves, which the DM cannot possibly account for, or present. It is generally perceived, for instance, that if the party is on a particular street, that every building on that street must have a ready-made account of who lives there, how they fit into the campaign, what adventure hook they offer and so on and so forth ... BEFORE the party makes a choice about what to do. When blowing this up to a world-wide scale, it makes it look like a sandbox is insanely impossible.
A sandbox is nothing of the sort.
It help to think of old movie lots where most of the buildings were cleverly designed facings, which had no actual contents behind them. The windows might appear to have wares, people might come out through the doors or lean against the frontages, but it was all camera angles and carefully concealed falsehoods. Building half frames is much cheaper and saves a lot of time.
In a movie, obviously, we know the actors aren't going into the bank or the general store, so they can be designed solely for how they will appear on camera as the lead walks by. The question arises, how do we know the players in a campaign won't do otherwise?
Perhaps it is giving away too much of a secret, but the principles lie in the same presentation as that offered by an illusionist or other form of magician. You raise the left hand, you wave the hell out of it, you draw the player's attentions to what you going to do with the left hand, and then you hit them with the trick. You don't let them walk into the bank because something really, really interesting happens on the street that draws their attention.
There are a couple of reasons this works. To begin with, chances are the players are not invested emotionally by the prospect of walking into a bank just to see what is there. Chances are, the players in your half-made town are only kind of wandering around seeing what there is to see, and as such are not terribly certain that the bank is going to be all that interesting anyway. If something interesting happens on the street, the player is likely to think, better the interesting thing I can see, than the maybe-probably-boring not-definitely-interesting thing I can't see. And in any case, whatever might be in the bank can wait. Right now, this terrible accident/fight/mud-throwing or pie-throwing contest happening in the street right now seems to be something they really ought to catch while its ON.
The magician knows the audience will be looking at the stage, because the stage is where stuff happens. He or she can count on the audience not excessively looking about them at the ceiling, or the drapery, or the other people ... and just in case they feel compelled to do so, the lights are brought down to focus attention on the stage. When you are running a sandbox, you have to learn to do the same thing. Yes, the whole street IS available for examination, but most players really just want something of interest to occur that will take up their attention for the moment.
Another thing the magician knows is the audience's habits. Knowing how people will respond to certain stimuli is of HUGE importance in running as a DM. Reading a few texts of psychology, or expanding your knowledge of human behavior, and testing that knowledge, will do more to improve your dungeon mastering than a hundred half-thought-out treatises describing being a better DM. Human beings are in fact very predictable. Magicians and other sales-pitch artists KNOW this. You can, with a little talent and patience, learn to make people jump and change their minds by feeding them exactly the right information ... and D&D is tailor-made for inventing information designed to misdirect or correctly direct your players. The hand waving can get immensely enticing, and if you wave that sucker just right, you can get your players to fall for it every time. Why? Because they WANT to fall for it. Falling for it is FUN. That's what magicians know. Audiences are willing, happy, cheerful dupes who want to be duped. It's what they pay for.
Of course there's the cynic in the audience trying to figure out how its done. And your players will usually include at least one cynic who raises an eyebrow every time a cart falls over, or an NPC staggers in the door bleeding from a wound, or a child stumbled through the scene calling for its mother. A cynic only does this because he or she believes they have seen it all, and that they can 'guess' at what's coming next. So the sniff and snort at the pie-throwing in the street and stomp into the bank anyway.
What's really funny about this behavior is that it, too, can be predicted. Once you know the player is a cynic, and once he or she has made it clear that they WON'T be fooled, you can lay the trap for them again and again, only with a little more subtle handwaving than before. A very good magician knows that the cynic in the audience is intensely looking for anything that looks out of place in the scene ... and so all the magician has to do it hang a second very subtle oddity on some part of their person, to draw the cynic's attention. "Hah!" says the cynic, "It isn't the hand, it's the piece of metal on their hat! Look at it! Can't you see it?" And then the cynic spends the next three weeks trying to figure out how that metal was absolutely necessary to making the trick work. When of course it wasn't. At all.
Really. Penn & Teller are absolute masters at this sort of thing. They usually have four or five fake elements in view, to screw with your damn head.
As a DM, you're probably not going to be Penn & Teller, but you will do very well to have something in the bank that is also 'interesting' and attention-drawing, which could easily be in the tavern, or the inn, or in a back alley, or anywhere at all in fact that the cynic insists on going anyway.
It isn't necessary to have a different something in each possible pathway - its only necessary to have something that can fit in ANY possible pathway ... that can be there to lead the player forward into some inviegled idea of your own making, which the player will cheerfully pursue because it sounds interesting.
The problem with railroading doesn't come up because you haven't got something that fits into every box ... the problem with railroading happens when you, the DM, haven't got a second trick up your sleeve - and you push and push the players to watch the one trick you have to show. A good magician should be full of tricks - and if the audience doesn't like this one, there should always be one more up the sleeve that can be slipped out while the hand is waving.