How do you prevent milling through tons of uninteresting play then?
Don't design an uninteresting game. Kinda funny, but its more or less the answer. Or at least a answer, anyway.
Ive said before that games, even stripped down to their bare mechanics, should be compelling and fun.
Many times you'll still need some additional layers of content to make for a compelling experience, but conventional narratives aren't exactly crucial at this level, and examining how you can make the stripped game more fun pays dividends when you pull that more bespoke interesting stuff back in.
For example, rock/paper/scissors isn't terribly compelling, but as noted, the dynamic it produces is similar to a conventional narrative. You have an inciting incident, a rising tension, a climax, and a denouement (and even sequel bait).
So right off the bat, r/p/s doesn't need much to be more compelling, nor to provide a very rudimentary narrative through play. If we shift up to something like very basic Pokemon battles, adding an assumed constant damage (because status moves and all that is an even more advanced step) and then the Typing system, we get a form of RPS thats dramatically more compelling.
So much so its still satisfying even as an adult to just play Pokemon like when we were kids, just spamming our best attacks.
And thats all without considering any of the content trappings; the context that you're battling your science fiction/fantasy pets, which makes the overall dynamic, and the resulting narrative, that much more compelling.
And as we add more systems and content, that experience, and the narratives it produces, become that much more compelling. Competitive Pokemon is a lot of fun to watch as a result, as are Nuzlockes and Speed Runs and all that, and when done well, even authored narrative experiences are a compelling integration. Arceus, for example, does a really fine job of integrating the classic experience with a compelling, authored narrative. (Jurys out on if any of the mainline games did it too; I believe 5th gen was the popular one for that specifically)
Neither are conventional narratives, but they are really compelling, and become even more so with the meta narrative that often accompanies these games.
Hence why game streamers have such a consistent audience; they wouldn't be near as popular, nor have taken off like they have, without the fundamental dynamics being produced as a result of their persona interacting with either one really compelling game or, as is the case with most streamers, a whole bunch of them. The parasocial nature of Critical Role has been criticized, but a lot of that is also just the meta narrative of how their table interacts with the game of DND that Matt puts on, which is quite compelling in its own right.
That is how do you figure out what process of play arises given certain mechanics and aesthetics?
Thats what playtesting and iteration is for, and why design patterns are such a useful shorthand.
Which does make sense. In other art forms, a lot of the process is the back and forth of consumption and design. When drawing, you do have to step back and look at whats being depicted. When writing, you do have to step back and read what you wrote. And so on. Revisions then follow.
Games are consumed by playing them, ergo playtest and iterate on, "revise", what was designed.
This is also why its important to have other people do this.
For example, when I was working on my combat system, my original idea was only 1 Action per turn, period. That was because I wanted the system to be really punchy, while also being much more indepth with the Actions themselves.
However, playtesting revealed that none of the punch, for the high octane high fantasy combat being depicted was lost upon allowing a 2nd Action, and in turn that nothing was lost by allowing technically infinite Reactions. (Practically its not a thing, but in some circumstances itll work out that way). This also, in turn, worked well to maintain my original goal of emphasizing input randomness.
BUT, recently, as in the past few days, I got distracted with a side project on trying to adapt the system to a Gun Fu style of combat, which by its nature is a lot more punchy and fast paced an aesthetic than the already bonkers high fantasy stuff I've been doing, particularly given I'm aiming for a good 50-50 on Gun to Fu.
So for the Gun Fu adaptation, playtesting has me figuring I'll be going back to the 1 Action per Turn idea (as that did feel better in testing), and I'll use the additional die that gets rolled to build up a new mechanic, that'll be about basically setting up your next move, and the overall dynamic, when interlinked with Momentum, should be something that feels similar to Gun Fu as we know it from Movies, but is fully interactive.
As for why the two felt different, thats just a matter of what aesthetically each take is presenting. The "fiction" if you will. While the high fantasy combat I'm going for is already pretty bonkers compared to the typical (casual dragon suplexing and all that), something about it just works better with a slower pace, and hence, more Actions being thrown around.
I'd kind of liken that difference to being the difference between a fight in Pacific Rim and a fight in, say, Extraction. More weighty and deliberate versus lighter and chaotic.
Suffice to say theres no real objective standard. We can try to quantify these things but its a matter of taste as to whether or not its what we want. One could easily flip the two ideas, and say the weighty, deliberate system wants only 1 Action whole the lighter, chaotic take wants more, if they aren't so much concerned with Pacing as I am and are more interested in controlling for some other aspect.