Players often notice that hours of dialogue choices produce an ending very close to the one they would have reached anyway. That convergence is a production constraint made visible.
Branches multiply faster than budgets
Every genuine branch doubles what follows it, so a sequence of independent choices produces an exponential number of distinct states the game would have to support.
Each state needs written, recorded and tested content, and most players will see only one of them, which makes the cost per hour of experienced content enormous.
No studio can fund that beyond a small depth, so branches are opened and then closed again rather than allowed to run.
Reconvergence is the standard structure
The common pattern lets a choice change how a scene plays, who says what and how characters react, and then returns the story to a shared state afterwards.
This preserves the feeling of consequence during the scene while keeping the number of downstream possibilities small enough to build.
Well-executed reconvergence is nearly invisible, because the differences the player noticed were real even though the destination was fixed.
State flags carry consequence forward cheaply
Rather than branching the story, games record decisions as variables and use them later to change small things: a character present, a line spoken, an option available.
This gives the impression of a world remembering the player while requiring only a small amount of additional content per remembered decision.
Endings are then assembled from those flags rather than authored separately, which is why ending sequences often read as a summary of choices rather than a distinct scene.
Voice acting hardened the constraint
Fully voiced dialogue means every unused branch has been recorded and paid for, and adding a line after recording requires scheduling a performer again.
That has pushed many games towards fewer, larger choices rather than many small ones, since a large choice justifies its recording cost.
Games with limited voice work can afford far more branching, which is one reason text-heavy titles often feel more reactive than expensively produced ones.
Players value acknowledgement more than divergence
What players describe wanting is usually recognition rather than divergence: evidence that the game noticed a decision, not proof that the world took a different shape.
A character referring to something done ten hours earlier delivers most of the satisfaction at a fraction of the cost of an actual alternative path.
Designers therefore spend their branching budget on visible acknowledgement and reserve genuine divergence for a small number of moments where it will be remembered.