I'm surprised people don't simply roll on random tables multiple times and then combine all of the results together when they're planning out dungeons or other adventure sites. Or maybe they do and simply don't talk about it much. As a low-prep referee who's often short on time, I do this pretty frequently and I find that it's actually more useful than just rolling once. The main thing to doing it successfully and in a way that actually eases the amount of work you have to do is to figure out the relations between the different results. Fortunately, we have the ubiquitous reaction roll table to assist in this.
For example, say you're repopulating a dungeon using a random encounter table. You're rolling to find out what monsters have moved into the depopulated area, and you're rolling a couple of times for each room for whatever reason (possibly because I suggested doing so in the linked post). You end up with 1d4 orcs and 2d8 slimes or something. Want to get an idea of why they're both in the room? Make a reaction roll. If the result is hostile then they're fighting, if it's friendly then they're allied (perhaps the orcs have tamed the slimes), etc. If you use the kind of random encounter table that I do, where you end up with a bunch of non-monster results most of the time, you can still use the reaction roll to figure things out.
e.g. Say you roll up a room with orc spoor and slime traces, and a reaction roll of hostility. Clearly, the orcs and slimes fought in this room, with one or more dead, slime-covered orcs in the corner. Neutral reaction? Clearly the orcs and slimes are not running into one another all that often - perhaps the slimes are active in the day, while the orcs come out at night? Or perhaps the slimes are eating the spoor the orcs leave behind - a half-dissolved boot or a leg of chicken stripped bare covered in goo would be a neat piece of garbage to find (it certainly beats the usual "wooden flinders").
This generates not only the encounters, but some of the dungeon trappings too (the truly lazy referee will of course, pull out the AD&D 1e and randomly generate the traces and spoor etc. that any given monster leaves behind by rolling on the bric-a-brac and weird smells tables).
If you recursively iterate a simple process like this often enough and record the results, the results eventually resemble a complex and dense set of relations between all the various pieces, even though chance is doing most of the work for you. Usually, you only have to do 2-4 recursions before it gets more complicated than most people can easily hold in their heads, which is also about the point that it starts to seem like a "world in motion".
I think many players of older adventure games or retroclones of them are probably familiar with the idea of doing something similar to this for treasure, where you calculate the total value of a treasure hoard and then roll for the percentage chance of magic items and other special treasures for various subdivisions (i.e. "5% chance of a magic sword for every 1,000 gp in the treasure hoard's total value"). But that's just a simple iteration of a process without recursion, and the relationships between the various treasure items is straightforward ( simple addition of the items to the hoard, or the replacement of a subdivision of the treasure by the special item), rather than recursive. The recursion comes about through generating and defining the relationships between the results.
If you're looking for a random table of possible relationships, especially if you are as lazy-busy as I am and wish to automate even this process, here's a random table of possible recursive relationships between multiple results on random tables:
1d4 Ontological Relationships between Randomly Generated Entries for Lazy DMs
1) Palimpsestic - The previous result is effaced except for a few traces (the slimes have eaten the orcs, only orc bones and treasure remain)
2) Additive - The previous result remains and the new result is simply added onto it (the slimes and the orcs are hanging out)
3) Combined - The new result and the previous result are combined into a single entity (the slimes are orc-shaped, or the orcs are covered in intelligent slimes)
4) Conditional - One or more of the results must be brought into the shared fiction via some trigger (the slimes are in jars, and if you're sloppy when you fight the orcs you will break them and release them)
You can apply recursion in all sorts of situations, not just encounter tables. Location generation is another good place to use it, with each result working as an archaeological layer of the building. Roll 1d4 for the number of archaeological layers in each sub-area, and then 1d4 for the number of significant digits separating each layer (i.e. a result of 4 is thousands of years between one use and another). This should also give you an idea of the relative time of construction.
Anyhow, as I said, I'm sure people are already doing things like this, I just thought I'd lay it out for anyone who hadn't thought of it, and to solicit suggestions from people who have even better versions of this sort of process that they're using.