There are these points in video games that simulate role-playing games where play often stops.
The game starts, and it is often linear for a bit, you are underpowered and have few options on where to go. You progress until you find an area where you can become more powerful, and then you acquire the means to travel to many different places, and then. . . .
Why? Why does it stop at this point?
Because once your linear goals are achieved, once you have the freedom to choose one task of many to prepare and address, that accomplishing of that task becomes work! Once the characters reach the 'domain state' and they are attempting their goals, those goals (calculate gold, find resources, wait months) feel like work, are accomplished at the table in a manner similar to the way work is accomplished, and gosh, aren't very exciting!
And don't we all have something better to do with our time?
I've seen it happen again and again at the table. When the characters finally get some footing, gain some levels, reach the point where their options are no longer constrained by 'what they must deal with' but instead they can address options that they choose, talk rolls around to another campaign. "We're going to put this aside for a while; We'll come back to it later."
And we all know how often that happens.
So what is the solution?
I think it's pretty simple. Allow the player to direct play, but do not let play become directed by the players.
Criticisms of sandboxes are many, but a primary one is that stasisticity. (I'll make up any damn word I want, thank you very much.) How to avoid? Constantly be looking at the choices of the players in that sandbox, and constructing activities (with clearly delineated steps) that occur from those choices. Then create dynamic sessions that don't involve the players picking from a bunch of choices that just feel like work.
This means that within these dynamic encounters you have things that further player goals organically. This means that the accomplishment of these objectives should be as smooth and automated as possible except for the relevant exciting quest-focused game-play.
Here is an example in my current game:
I have a player who is a dragon wizard. He wants to learn dragon magic. He is fifth level. He can no longer advance along this path of dragon magic until he learns the language of dragons.
I could have presented this fact to the player and then waited for him to come up with a way to try and find out how to learn this language. I could have even presented options, such as 'go ask a sage' or 'ask around in the bigger city' and let him try to solve the problem himself.
This is what I did.
He had recently used his powers to kill an ogre. He then woke up with the ogre haunting him, preventing him from getting any sleep. He then consulted a local priest and then the group specialist on the dead (a henchmen who's class is Deathmaster) to discover that he would have to either bind the spirit to a nearby creature or find a much more powerful sage to banish the spirit.
Conveniently, since they had a deathmaster, a powerful ring of undead utility and a dumb flying lizard in a cage.
The player was presented with two options immediate options to meet his goals. Sever and bind the spirit in the dragon creating a true dragon familiar, that could teach him draconic at the cost of a character level (required to gain the powers of a familiar), or pay gold to a powerful spellcaster to banish the spirit and go on the adventure to find someone to teach him draconic. Of course, he could have investigated in his own way how to learn draconic himself!His choice, most assuredly did not feel like work.