There's a column on the Escapist that I've been following pretty much since it started appearing there: Check for Traps. The guy behind it, Alexander Macris, has some really neat ideas about running a game. I'm currently (very slowly) designing a campaign using a lot of his advice. There is one of his ideas, though, that I think may be a little more complex than he's made it out to be.
This idea is his Agency Theory of Fun. Don't worry, I won't make you follow the link if you don't want to. This theory is Macris' answer to the prevailing opinion that the Gamemaster's job is to make sure everyone has fun. He argues that many GMs are afraid to give their players real choices because, if things go poorly, it could ruin their players' fun. But that has the unforeseen consequence of making none of the players' choices matter, which isn't fun.
In its place, he states the GM should prioritize their role as a judge, arbitrating the rules of the game fairly and consistently. By risking the various catastrophes a bad roll can cause, the players' choices have meaning, and making meaningful choices leads to fun.
Frankly, this is the type of game I love to play in, but it isn't for everyone. I've met some people who would absolutely hate this style of game. And there are reasons beyond fun maintenance why the other type of game gets run.
First off, trying to run a game in which the players truly have agency is hard. A lot harder than one in which dice rolls are fudged and the game is gently railroaded. In a game where the players are allowed to do whatever the rules and the setting allow them to do, the GM has to either plan for every eventuality or be very good at improvising. Rather than just creating the NPCs and settings and plot along the carefully planned out path, they have to have everything built in every direction or be caught with nothing when the players strike out for the blank parts of the map. Not to mention, a game with some direction can create some really great stories.
And, as for the players, some people get really attached to their characters and, by extension, all the gear, companions, and voluminous backstory that comes with any PC that has existed for more than 20 seconds. A lot of effort goes into these things and it can be a blow to have some or all of this taken away by chance or a bad decision. Even a railroad plot has its fans; some players have come to respect and expect subtle clues to follow to move along through the GMs plot.
As much as people rail against railroading, it can serve a purpose. It acts as a safety net, both for the GM and for the players. It protects things that are important to the group, like characters, gear, and a coherent plot, at the expense of things that are - to this hypothetical group of gamers - less important, like agency. Player agency games aren't for everyone. Railroaded games aren't for everyone.
The trick is knowing which game you're getting into. I think that 99% of complaints about either game style come from people who found themselves expecting a game to be run (or played) one way and find that their GM (or players) are intent on treating it like the other. It isn't that the other guy is a bad roleplayer, it's that each group is trying to play a different game.
When a group of gamers starts a new campaign (even if it's only a one-shot game) a kind of social contract is formed. The GM and the players agree to abide by a series of unwritten rules. How player agency is handled in a game is a big part of these rules. The problem, from my point of view, is that most times, these things aren't discussed, so everyone goes into the game with a slightly different idea of what the rules are. So, it isn't long before someone breaks the social contract and everyone stops having fun.
I'm not saying that an actual contract or set of rules need to be decided and debated before everyone sits down and rolls up characters. I do think, though, that people need to be more aware of these essential differences in play styles and should make sure that they know beforehand what kind of game they're in for.
The GM especially should keep this in mind when telling his prospective players about the game he's planning. Right after system and setting, it wouldn't hurt to mention things that give the players some idea of what they're walking into. Even little things, like saying, "I've got this great epic story all mapped out and you guys are going to be at the center of it," or "Just so you know, the kill switch is going to be on from game one, so don't expect me to pull any punches," will give players an idea of how much railroading or agency they can expect.
Player agency is an idea that ties into a whole lot of other parts of a game. The more everyone's on the same page, the more fun everyone will have.