Uh, it make perfect sense to have multiple/different outcomes in games where choices got consequences and matters though.
Sure, but locking you out of content because of those choices is really dumb. I'm not saying consequences shouldn't exists, but rather that consequences shouldn't have you going back and forth between text, because that's not interesting and it's just redundant. In most other video game genres I would agree, for example, I like how Shadow the Hedgehog have you backtracking the level in some cases, but that's just not really viable in kinetic VNs. I guess in sand boxes is different, but even then, my problem is that the gameplay revolves around reading, so instead of having you explore things as you would in any other kind of game, you do it so by reading, and, as said, that's not all that interesting and can get repetitive.
Since we are talking about games like Milfy City, then it applies even more. Basically, I think there should be a balance between backtracking for hidden or different stuff and the time you spend on backtracking, and seeing how these games are mostly about reading, then "backtracking" hardly applies and it's most of a waste of time than anything, BUT a completely linear game sounds boring, as you stated.
My point is, as stated before, that, say, an "Easy mode" would basically be a self-implemented walkthrough, with clear descriptions of what each option do, which one is better, and so on and so forth. And even then, a normal mode should have enough information for you to figure out what is going to happen without needing a walkthrough or trial and error. That means that the second time you are going to play you can know exactly where you went wrong, and retry.
These kind of choices are really hard to balance, but see it as in RPGs: When you get new weapons, you are going to compare them with the old weapons you had, and you may not choose the "best" weapon, but the one you may like the most. Applied to VNs, if you are given at least who that choice will affect, then it would be easier to the player to choose WHAT THEY WANT, not what is empirically "better" or "worse".
Baseball, hockey, footfall, soccer... bad game design
Those are not video games, and your analogy is... Weird, to say the least.. Practice makes perfect, but the thing is, if you do like a book, which is the most closer example to a VN, you will read it again without having to rely on gimmicks. You don't need to "practice reading a book", and the only thing you can improve is your reading comprehension.
Even then, as an sportsman, regardless of your sport, what you are doing is "min-maxing" your options. It's not just a matter of "I do this blindly over and over and I get better", you need the right technique, a good coach, and then minimize the time you spend trying to better yourself while maximizing the gains of that action. If you don't, then you indeed have a bad coach or a bad exercise routine, and that can equal to bad game design.
My problem with the term "replayability" is something more personal. Personally, if I finish a game once, and that includes finishing all the endings, I expect not having to go back to do menial stuff
And I think my best example for this is Banjo-Kazooie compared to Mario 64, actually.. That would just be extending the game time artificially.
"You completed the game! But there are exclusive bad ending CGs! So you have to lose on purpose in order to get 100%!", I think that sounds just terrible. If something, there are games that allow you to 100% the game is you finish the best ending, and that includes giving you the "bad endings" CG, if there are some in the first place.