I tried to choose three very different types of nonlinear narrative. One is a story in a choose-your-own-adventure format, the second a mystery presented in the form of a police file of evidence and information, and the third a live-action role-playing game.
The lexia of this story are all contained in a single HTML page, with section labels used to move between them. The lexia are very small, certainly less than screen-sized on any browser, perhaps too small at times. This story is structured much like a choose-your-own-adventure, making the structure and methods well known at least to me. The setting and situation are strange, though, and no background or introduction is given. This may be the author's intent, since the story seems to describe a dream, or even a dream within a dream.
There is a decent amount of branching in the story, though the various branches often return to common elements. Examples of this include one lexia which is the root of a tree of choices all of which lead to the same ending, and another where two choices (the second two) give the same result. The story is unfinished, so I'm not aware of the total extent or conclusion of the story.
The division of the evidence and facts of the case was logical for a police file which had been dredged up, though it could have used more cross-referencing, such as links to relevant evidence or reports on the pages for witnesses. The lack of a way to take a linear sequence through at least the basic information made the ramp-up difficult. As a reader I had to spend some time exploring the hierarchy of the pages for the case before deciding where to start reading. The individual information was segmented in the natural way, rather than optimized for web viewing, such that an entire interview, news clipping, etc. was on a page. This worked well.
here was some information which was duplicated in several places. This is a fine thing, but unfortunately the information was often incomplete in one place or another, in some cases enough so to be contradictory. For instance, the alibi chart was a useful collection of data, but it listed a meaningfully different alibi for Carl Rose than the one in his interview. The reader had to dig a bit to find the right bit of information, and the static online nature of the case info made it frustrating to discover that a piece of information the reader wanted didn't exist. This was partially a case of omissions of what was probably expected to be irrelevant information, but it detracted from the realism of the crime report.
Lifeboat is an Assassin Game (LARP), a very different type of interactive narrative from those presented on the web, but also a form very familiar to me personally. Its author is also a friend of mine. Different points of view are one of the most important narrative tools in an Assassin Game, as the game setup only works when each player knows (through his Charactersheet, Bluesheets, Greensheets, etc.) the things about the game world that his character should know, as well as how his character would be likely to react in the unexpected situations which arise in the game. I know from experience that the distribution of information between characters and groups in a game can be a tricky business to balance correctly. Since Lifeboat has only 6 characters, and no groups meaningful enough to have their own sheets, that problem is a bit simpler here than it would be in, say, a ten-day game with 83 characters, about 50 groups, and far too many game-mechanics (not that I'd ever be involved in one of those :).
One thing this game does which many don't is attempt to present all of the information that a GM needs to run the game in a coherent, easy-to-read collection (linked to from the GM Info section of the page). Most games are written and run by the same people, and are only run a limited numbe of times and then put to rest. As a result, though all GM teams try to avoid it, inevitably much of the information needed to run the games turns out to be only in the heads of the GMs, rather than written anywhere. This leads to difficulties both during the game if the GM team is not all present, and in ever running the game again in the future. Lifeboat does a good job of presenting all of the needed information, and doing so in a logically separated way (separated by content, not by lexia sized for web viewing, since the web is not the primary form of presentation of the game), though again the job is significantly easier for such a small and simple game.