[Bruno]

Regarding forking paths:

Choices for the players to make are good, but they have to be meaningful choices. Having dead-end paths doesn't make for real choices, unless there's a consequence for taking the dead-end (good OR bad). If the players know there's a chance to be surprised and cornered by a wandering monster, or wasting precious time before the virgin sacrifice, etc these are all meaningful, as is a chance to find more treasure, especially a weapon useful against "the end boss".

Similarly, having two paths to the same end area isn't really a choice unless there's a notable difference between them. If path A) leads through two dungeon rooms with two "equal level" encounters and path B) leads through two dungeon rooms with two "equal level" encounters, that's not really significant. But, if path B) has NO monsters at all and offers a crazy good shortcut, and the players have any way of determining that, it's not really a "choice" either - you'd be stupid not to take the fast safe route.

It's like if you have a "choice" between two advantages that cost 5 points - one gives +1 to your favorite attacks, and the other gives +2 to your favorite attacks. That's not a choice at all, that's just a booby-trap for someone who doesn't notice the second advantage. Eidedic Memory in 3e was like that - it was either insanely good (and you were stupid not to take it) or insanely bad (and you were stupid to take it).

If you can't come up with a significant choice, and it's a small area, straight-through is perfectly fine. Having a scattering of small 3-5 room "straight-through" dungeons works surprisingly well, because the player choice is in which dungeon to poke their noses into (and when) rather than having different routes through a single dungeon.

The classic Keep on the Borderlands was based around this model, although the "areas" are larger because the original game played through pretty darn quickly (either you die immediately, or the monsters die immediately :)