When thinking about games that attempt to model the world, the first thing that comes to mind is "book keeping." This was often a bane in the past, and probably still is, but one thing that I think might help is the proliferation of electronic tools. In general, I actually think VTTs can help bring back crunchy, world modeling systems because a lot of the niggling details that could be troublesome to work out at the table can be handled by the computer. The most basic example, which is I think a thing that is essential for these kinds of games, is inventory management, including encumbrance. The VTT or D&DB or whatever can keep track of how many days' rations you have left and how much they weigh. That's great.
Something to learn from video game designers is that trying to model actual physics is a dead end outside of specific and tightly focused simulations, and even then. The truest to life flight sims aren't really all that close and really can't be without it becoming a barrier to play on many different fronts; there's still a great deal of abstractions going on to ensure the game works as a game.
Instead whats preferable is something some call sweet lies; ie, abstractions that look and feel the part but simplify and reduce the complexity and induced constraints to model them. When you have a magnetism power in Breath of the Wild, you aren't trying to build actual magnetism physics. That'd be impossible to run. Instead you just add tags to the right objects and the power only works on those.
Something else to learn though is that any sort of items you introduce to the game
should reinforce the gameplay loop. This is less important in video games as they can easily model infinite amounts of clutter with little overhead, but not so much in TTRPGs where trying to directly model clutter just bloats the system. Clutter though isn't just random buckets and trash; its also random parts of the street or buildings a Superhero might be inclined to pick up and throw at the Supervillain (or vice versa).
Rather than trying to model these types of clutter, and all the infinite varities inbetween them, you would typically just define how whatever character mechanic interacts with these, typically Strength, relates to the options open to players. Ie, 5 Strength says you can barely lift a sword, 30 says you can rip off half the building and throw it onehanded. It doesn't actually have to be Strength, mind, but its all the same underlying design principle.
The main concern is getting the scaling right, and being explicit about what different levels in that scale do.
But an often unnoted concern, at least as far as doing this in ttrpgs, is embracing emergent gameplay in these mechanics. Some games put in a lot of wonky exceptions into their rules that try to plug these emergent behaviors (or worse yet design them in such a way that emergence isn't possible at all, or is just super weak), and most don't give any guidance to GMs on how to handle when these things happen and how to lean into them for great effect.
The peasant cannon, while it doesn't actually work regardless, ultimately only deals 1d4 damage per 5es ruleset. If we get rid of the nonsense of the cannon, and just assume some other more typical means of launching an object to try and harm something, its kind of pathetic that it doesn't do much, isn't it?
It makes environmental interaction not very worth it to explore which in turn means gameplay is going to narrow in a lot of areas other than just this one part, because now even if theres spots where environmental interaction might be beneficial, people won't make use of it because one option is basically a waste of time.
In regard to modeling the world, one thing that I haven't seen very much of (though instances of it are out there) is methods of generating relatively large-scale events in order to portray a dynamic world, i.e. things are happening even if the PCs don't interact with them.
You mostly get this as "random events" at domain-level play, where the GM will roll on a table for something to happen, either as part of the PCs' fief or abroad. But it applies to a lot of other things, such as random weather tables (which can be a lot of fun if used correctly, albeit more so in low-level/magic games), or even an economic fluctuation system I once saw. In those cases, I think that rather than rolling them at the table and letting things unfold in the moment, it works better if the GM makes those determinations ahead of time, and extrapolates where necessary to help tie things together, creating backdrops that the PCs can subsequently react to if they so choose.
Now, that requires a lot more work by the GM, work that needs to be done prior to game, but it can result in a very immersive play experience, and it's one aspect of "generating the world" that I think gets overlooked a lot.
A method I advocate for for this is to essentially write timeline flexible stories that assume the total absence of the players. Ie, if they never interact with that story's elements, the story plays out as written.
But by making them timelime flexible, meaning you can time certain acts within that story to happen when you need them to, then you open up the possibility of PCs intervening, and then it becomes a matter of working out, either on the fly or session to session, how the PCs interactions change the written story, which becomes dramatically easier when you have a fully functioning and finished story to work with; you're not worried about how the story will end while also trying to figure out how to incorporate whatever bonkers things the PCs did; unless it was so bonkers that it unravelled the entire story.
But that then becomes a matter of improvising a new story, which outside of specific circumstances, might not need to happen in-session. GMs can and should call a session early if their material got blown out of the water, and most of the time its unlikely you're going to be in a situation where you have to immediately provide new story beats. These games
do have downtime baked right in, after all.
(though instead of flight speed, it's often to determine the vision capabilities of various creatures before the Druid shapeshifts to go scouting and-or to do stuff at night)
