Web-annotations theory
Is this the first 'annotations wiki'?
In any case, there will be many more, so it will be helpful to thrash out some guiding principles...
How much detail?
Aspiring novelists will see the annotation-process as a form of apprenticeship, reverse-engineering Pynchon's bag of tricks.
Others may see it as a course of advanced study, hitchhiking on Pynchon's enthusiasms.
Pynchon seems to go to enormous lengths finding non-obvious research sources, but by now the Web is huge enough that most factoids can be verified via linkable Web resources.
Most all novels blend the fictional layer with a factual background, and sifting one from the other seems a legitimate annotators' task. Similarly, checking for anachronisms of history or language. (Historical characters' actual ages at the time of their appearance should be noted.)
By including lots of links to detailed resources, the notes themselves can be kept compact. (Longer thoughts belong on a separate page.)
Useful resources include maps, pictures, etexts, etymologies, pronunciations, etc etc etc.
Handling spoilers
The spoiler problem is vast and problematic.
In Ulysses there are so many baffling references that are (more-or-less) clarified later, that trying to avoid spoilers is hopeless.
There's a possible trick of hiding them with white-on-white text, visible only when highlighted (via mouse or keyboard). Can this be done with the current wiki software?
In the long run, most of those who use these annotations will either be re-readers for whom spoilers are a non-issue, or novices who will give up early, so minor spoilers are not a big deal.
Specific test-cases with AtD: a celebrity cameo that's fun to spot and widely spoiled by reviews that mention him; a minor character who later becomes major (does it hurt to warn people to pay attention?); a character who dies suddenly while contributing an important early section-title (does it hurt to reveal that early death?).
Also, if spoilers are to be avoided, what about Gravity's-Rainbow spoilers?