Sorry for posting in a week old thread, this took several days to make.
As this is a huge post full of all kinds of ideas, I’ve ordered it into three parts for readability.
While much of what I’ve written is in reaction to the previous posts, I have elected to not quote any of them. Instead, I’ve structured the post in an “article” style. This is largely due to the length and how the parts flow together. Think of it as a compilation of my own thoughts and the good ideas I’ve stolen from the thread. 
Hopefully it won’t end up too pretentious or obvious.
And, of course, as I’ve yet to complete my work everything I say is based on my design philosophy rather than practical experience. We’ll see how I feel after I’ve actually finished it.
The point of interactive fiction:
To me, what makes interactive fiction so interesting is “what if”. If there is a choice, then I’d like it to have some function. Whether it is to give flavor to text that personalizes the situation to the reader, or whether it’s an actual choice that alters the story greatly doesn’t really matter that much.
What I don’t like is the idea of an “illusory choice”, as it implies that what you have chosen doesn’t matter. Now, it’s fine that it might not matter on the bigger scale(doesn’t branch), but it definitely should matter to the reader. What I mean is that it should make the experience greater in some way, for example by the aforementioned flavorful descriptions available only to the particular playthrough of the story. To call it an illusion of choice is a mindset that might lead to a less-than-optimal outcome. If you feel that the story will only work if forced along a certain path that’s fine, as long as the path looks personal. And it’s the small things that make it feel personal, not the large things. By necessity, the larger branches are static to themselves, while tuning a small line here and there gives it the charm that is the point of interactive fiction. It might be a good idea to focus on the small stuff more, rather than on the branches. If the story divides to paths with large differences, then a lot of effort must be expended on them and it leaves less time for the smaller details.
What the point of this is, is that if you railroad the story, then you have to frame it in such a way that makes it clear that there is no choice. It’s a deep insult to the reader to pretend there is a choice to be made when there isn’t. If you, as the creator, don’t think of it as an illusory choice, but rather as an opportunity to embed flavor to a particular scene, then your mindset is far more productive. Instead of “fooling” the reader, you’re making them enjoy it more. Thus, I think the focus should be more on the “what can I add as interesting description” rather than “how can I make it seem like this had a point”.
To me the illusion of choice is the only thing that can break interactive fiction. Once the player/reader realizes that the choices are an illusion, their enjoyment plummets. They can’t trust the author anymore, which leads to disappointment. Yet on the author’s part it’s unfeasible to expend a vast amount of work on endless branching. What to do, then?
Don’t pretend that a choice is meaningful, make the consequences meaningful. A meaningful consequence can be as simple as a sentence of flavorful description. That a specific player gets it in their playthrough makes it “their” playthrough if it’s done properly. They feel that they have experienced a personal story and it’s special to them. But the description has to be meaningful in the context and has to arise from the choice that has been made earlier or at that moment. It has to give more point to the story. To call it an illusion is, in my opinion, to not utilize the most important aspect of the idea of interactive fiction. Which, to me, is a personal experience.
Practically this means that having choices that result in different paths isn’t the most important part. What’s most important is for it to feel like it’s your unique playthrough. Consequences of choices are important, not macro-consequences, but rather micro-consequences. The great forks in the road are necessary, certainly, but they mean nothing if the story isn’t peppered with tiny bits that actually make it feel like it’s yours. This is what makes interactive fiction different from computer games. Many computer games have forking paths that offer different mechanics but what they lack is the “personalization” that the best interactive stories offer. A game, by necessity, focuses on the gameplay aspects and level design, while interactive fiction focuses on the writing. This is especially true with medium-level branching(as an example in a computer game, the level itself is the same but the context and the monsters you fight change), as it’s a matter of re-useability of resources. Interactive fiction doesn’t have the same costs in creating resources. Sure, it saves a lot of effort to recycle some things, but it’s nowhere near the level of creating new animated assets and the like in a computer game. Besides, blindly recycling writing will often lead to jarring inconsistencies which will bother the reader.
To use Tin Star as an example, the moments that I remember the most are the small things, like taking Carson along with you to the dam, where he comments a line or two, and the dam people offer him a drink that he refuses. I had chosen to take Carson with me. I could have taken someone else. The primary point of the scene itself doesn’t change in any way, so Carson doesn’t affect the branching as such at all. But his presence is what makes it feel personal. Not the big parts, but the tiny details. Tin Star has very few true forks in its story as far as I’m concerned, but it does the most important thing right, makes the first playthrough feel special. The ones after the first destroy the illusion of choice and lay it bare, but it doesn’t destroy the tiny flavorful parts that are different based on the things you do. And as far as I understood, many(if not most) people play through a title only once.
It’s also good to note that adding these micro-consequences doesn’t really take all that much work in the end. Let’s say each scene would have from one to ten additional, different lines depending on a previous choice(or several of them). Not really that much compared to the meat of a scene, yet it adds so much character to a scene that it’s unreal.
It’s good to keep in mind the following: The point isn’t to make the micro-consequences be branches or interactable, just a set of “if” clauses that activate if some boolean is true or false. This limits the work needed to implement them, so you won’t be swamped by them as you would by a branch.
That said, I do believe that a proper interactive fiction story has to offer macro level choices and consequences too. If the experience is more or less the same with two different playthroughs, it’s a failed execution. What’s the point? It’s not interactive if you can’t affect it. Failure is fine if it’s framed properly, but a story that relies on the failure of the protagonist just because it’s the most common way of creating drama is a waste. For once there is an option to both fail and succeed, why not make the most out of it? If you want a static story, why spend the extra effort to make it interactive fiction? Obviously there are many reasons, but I’m trying to make the point that it’s good to utilize the strengths of the medium. Still, in the next part I’ll address why it’s not necessarily a problem to have forced failure if framed properly.
Regarding plot-mandated failure or success:
I don’t think loss has to happen for a story to have impact. This is especially true in interactive fiction, because crafting the whole is impossible. In normal fiction, everything has to have purpose and a meaningful point. Every win and loss is calculated to make the story greater. Interactive fiction, however, makes it practically impossible, because you simply can’t craft a scenario where there are both meaningful choices as well as meaningful structure. A single failure might ruin the story-experience of one player’s path(due to the choices they made), but could make another’s experience great. Thus, you can’t do the same “math” as you can with static fiction. Because you cannot(or rather, should not) control every variable, you can’t create the perfect story. Trying to do so is an inefficient use of time and can result in bitterness on the player’s part. Instead, you should create the best “ingredients” possible. The player’s imagination will fill in the necessary blanks, if they have enough food for thought. And giving them that possibility is what matters. Not in precisely crafted meals, but in quality ingredients that they choose and mix themselves with their choices and the consequences that arise from them.
As such I don’t think the basic structure of interactive fiction should be the same as with static fiction. I know that some people play interactive fiction games “to be the hero” or such, but I think that’s just the surface of the medium. The deeper parts offer so much more. And to get to those deeper parts, you might have to abandon the structures laid out by other mediums. As for what that actually means, I hope to explore it with my WIP. I’ll make it public once I feel confident enough in the basics, but that’s probably a month or five away. 
Still, anything can work. A story where there is only failure, or only success, can certainly be a good one, if the interaction is meaningful outside those parts of the story. Who are there to witness the failure of the character? How do they react to it? How does the player react to their reactions? That is what matters, not whether you fail or not in an arbitrary “Hero’s journey” style mathematical formula.
I offer this: If you feel that the MC has to lose, then don’t think too much about those parts, implement them as you feel you must. Instead, focus on the consequences of failing. How does the world change because of the failure? And more importantly, how have the MC’s prior actions framed the failure? If the player has chosen to interact with a wise mentor before failing, they might be very disappointed at the failure and express it, which will create an opportunity for angst. But if the player instead chose to interact with a nurturing healer who will take care of them after the inevitable failure, the resulting drama will be massively different. This way, you can keep the “main path” just the way you want, and make the “meat” of the interactivity in the things that happen before and after the failure.
As mentioned above, I believe that a player is most satisfied when the story feels personal, not when they feel successful. Often, success feels like they’ve affected a story(which makes it feel personal), but that isn’t all there is to it. You could make a story that consists only of bitter failures, but still make it satisfying due to character interaction and introspection.
Perhaps I am naive, but I tend to think that people who yearn for success tend to play competitive games where they can feel it, and people who want an interesting story read interactive fiction. And to me, these things we create aren’t games, but interactive fiction. To see them just as games is terribly limiting. The potential of the medium is unique and should be used for what it’s best suited for. Not simply to imitate games, but to create a new kind of experience that can only be achieved with digital pages, because they can be edited on the fly and thus they contain tiny differences that make all the difference. What would be a huge amount of animating and asset creating in a regular computer game is trivial to create with interactive fiction.
What I’m trying to say is that interactive fiction is as different from regular fiction and games as those things are from movies.
The medium is still in its infancy, so there is much to be learned. And by sticking too much with the older mediums you can’t go forward to squeeze the best out of it.
Difference between success and failure:
Success and failure need not be opposites as such, they just lead to different scenarios, the results of which are partly good, partly bad. Winning against a certain character could make them bitter and thus you might not be able to become their friend, or losing against them might make them friendly. Of course, such things need to be fairly apparent as the worst thing that can happen to a reading experience is a surprising result that you hate.
As I wrote above I think of interactive fiction as a story rather than as a game. And a story should never be a failure. Thus, the person experiencing the interactive fiction should never run into a game over or non-satisfying conclusion. A choice with a stat-check that leads to a “failure” if the stat is too low shouldn’t be seen as a failure, but as a different end result that leads to different opportunities. Every choice should ask “what direction do I want to take the narrative to?” rather than “how can I win?”.
As such, to me stats are a derivative of “what direction do I want to take the narrative to” rather then “how can I win”? If a player focuses on a certain stat, that implies that they enjoy the narrative that springs forth from that stat. Thus, the scenes that are more heavily related to that stat are gated beyond stat-checks relating to it. I don’t really like the “choose the stat you want to solve the situation with” type of “choices” since they have little meaning(sometimes they are important as micro-consequences for their flavor text, but often they aren’t really that well done). Rather, I like it if the stat leads to a small mini-scene relating to that stat. It creates content that is connected to it. If the player wants to play a skilled hacker, then a high hacking skill should lead to hacking related mini-scenes, rather than it just being a way to “solve a problem” even if a flavorful one.
Furthermore, one could consider that simply trying to use a certain skill to solve a problem would lead to it being weighed more heavily in the narrative, since that is clearly what the player wants. This means that even if the check is a failure and you haven’t done an interesting follow-up to it, in the future it might matter more. I guess this is sort of an argument for the “failure increases skills” design choice. Despite that, I’m personally not that fond of such a design choice, since my nature as a person is to maximize gains and I’d be drawn to those choices even if I didn’t exactly want them. It’s best if skills are increased during “intermissions” or the like. There should never be a pure choice between story content and skill increase. They should be separate.
In any case, to expand on this, think of the following. Rather than simply succeeding in a check/choice that results in something, I think it’s useful to sometimes consider the idea of a success being “setup”. Meaning, even if you inevitably lose against the villain early on, succeeding at sword-fighting will cause a wound to the villain. The story should remember this setup for the final battle against the villain, where the wound is referenced and you can then “cash in” the success you set-up earlier in the story, thus allowing for a dramatically satisfying conclusion to your “great at sword-fighting” narrative. For example, the villain could comment on the wound and the MCs improvement or there could be a flavorful description of the wound hindering the villain or such. A player who wasn’t good at sword-fighting earlier in the game wouldn’t get this setup and thus can’t cash it in. Instead, they’d have a different description, one where the villain would be caught unguard by their increased skill(should their skill be higher than before). This seems a lot like a delayed branch, doesn’t it? I’m guessing the only difference is in how it’s framed. Mainly, that succeeding or failing at an earlier choice can come into play later on in strange ways, rather than having an immediate effect on the branching or the flavor text. As a reader/player it’s very enjoyable to have a far-reaching effect on the story, for me at least, so such a thing would come as a pleasant surprise.
To conclude, the idea of what interactive fiction represents is the most important thing that the player and the author must agree on. If one of them wants it to be a game about winning while the other desires interesting narrative, it’ll end in sadness. A compromise is likely to be unsatisfying too. I believe there’s enough space for several different approaches, but the problem is how to get the reader and the author to be on the same page.
What I am sure of is that everyone agrees that meaningful consequences are what makes interactive fiction it’s on medium. In the end, isn’t a meaningful consequence one that brings joy to the player? It can be 20,000 word sub-path but it need not. It might be just a short sentence uttered by their favorite character at just the right point, one that brings a tear to their eye.