If you want to create a more elaborate table, the first thing you must dispense with is the flat, linear construction of "die total equals result." You see the table everywhere: roll a d6; 1 = bugbears, 2 = goblins, 3 = orcs ... and so on. And then, when you have completed the die roll, and the encounter, you go back to the same die roll to produce another encounter. And after you've played using a table like this for a long time, you get sick of tables and go looking for some other aspect of D&D in which to get interested.
Now I want the gentle reader to recognize that even a board game, the thing that most D&D players decry as the evil thing that D&D has superseded, is more complex that what the DTER (die-total-equals-result) methodology offers. Let us consider, because it is familiar, the game of Monopoly.
Let's postulate the car, the thimble and the iron sitting upon the GO square. Two dice are rolled for each. The car gets a four and a five and moves to Connecticut Avenue and buys it. The thimble gets a two and a three and moves to Reading Railroad and buys it. The iron rolls two threes, landing on Oriental Avenue - buying it - and then rolls again, getting a five and a two and ending up on States Avenue - which the iron also buys.
When the next turn starts, the pieces do not move back to the beginning and start again. No, they proceed from the point where they are at and new results occur. Each turn is a stepping stone towards the next turn; each previous turn creates a formula that changes the position of the pieces for the rest of the game ... and in this formula a multiplicity of games results that are unpredictable and therefore consistently interesting to people. Monopoly is still played as a game for this reason.
However, the DTER method does NOT provide this. When you roll the die that determines that you throw orcs at the party, DTER tables do NOT then provide a stepping stone for the next roll. The next roll occurs completely separate from the previous roll and for that reason the table is absolutely and completely useless for the construction of a sandbox world.
What would be the alternative? Well, to describe that, allow me to propose another example, also from the familiar file. Consider the game twister.
Here we have an extremely simple catalyst. We have four fundamental body parts: your left and right hands, and your left and right feet. And we have four colors: blue, yellow, red and green. When the arrow is spun, determining this particular body part must be applied to that particular color, we have 16 possibilities. This does not seem like very many ... but the game transcends the possibility in favor of something more important: the building up of result upon result. It isn't just that you put your left hand on the yellow circle, it is that then something will need to be done with your right hand, and then your right foot, and so on, while other people are also attempting to contort their bodies likewise.
If you are going to design tables for D&D you must think at least along the lines of Monopoly and Twister. How does this result lead to the next result? How does the combination of these results ultimately create new and unique sequences or troubles that are, at present, unexpected or unfamiliar?
In making a weather table, for instance, if I randomly produce a likelihood for high humidity; and I randomly produce a temperature; and I randomly produce the occurrence of a warm front or a cold front ... then how does the combination of these things produce OTHER results, such as fog, violent storms, high winds, a beautiful dry sunny day, and so on and so forth. To simplify it, if I roll three d6, they give me 16 possible results, balanced on a bell curve. But if I roll three d6, and the specific results of each of the three dice affect the result of the other two, then I produce 216 possible results. To make that more clear, if I designate the first die to be a type of vessel, and the second die to be a given weather result, and the third die to be a tactic for avoiding the oncoming vessel, then the combination of all three becomes a necessary interpretation of how A affects B which is in turn affected by C ... and not just designated possibilities. Do you see? The results are logical possibilties, given the perameters of the model you construct. If A lands at Connecticut, and B lands on Reading, and C lands on States, we have a model to guess at who is the most likely person to successfully obtain the Electric Company.
And if the next die 6 I add to this mix does NOT reset the game back at GO, I have 1,296 results to interpret. And the next die produces 7,776 results that can be interpreted.
Finally, if we then understand that the individual constructions which produce the interpreted results can then be balanced to fit a given circumstance, so that 3d6 can be used to determine the type of vessel, and a number of different dice and results can be used to determine the weather, and player choice can determine the tactic, we have an untallied number of results ... with possibilities worthy of making the game forever interesting.
No, obviously, it isn't easy. But the payoff is greater than crappy DTER tables can offer.