All The Tropes Wiki Project, Part VII

[Labster]

We will use at least some advertising (more for lurkers, less for active
tropers). ... But we definitely don't want to sell user
data.

Are there advertising networks out there that don't collect user data, or are you planning on doing direct sales?  I guess the good news is that ATT doesn't have to worry about its competitors providing a better user experience.

We've been there for years, but later, the wiki lost its fun. ... Then we learned that you had made the site rip, and downloaded it.

Tell me about it.  I have really done any troping in years.  The programming part keeps me interested though.  As for the site rip ... I was a much worse programmer back then.  I apologize about the quality of the dataset, with all of its duplicate pages and such.  I have improved much since that day, but still know nothing.  (Also, CSS Zen Garden looks fun, but also overdesigned for real websites.)

but then again, Wikipedia is also struggling with their WYSIWYG editor.

Less so than you'd think based on the press.  It was pretty shaky at launch, but it's improved quite a lot since then.  The real issue is that forcing change onto Wikipedians will always result in chaos.

Open Source is great, but in our special case, there's a problem: Our
software is highly specialized - it's great for a troping wiki, but not
for much else. The only ones who could put it to use are us - and the TV
Tropes crew. Just the folks we don't want to schlep for free for. You
understand our problem?

I do see your issues, but I reject the premise that it would only be useful for a troping wiki.  The central idea of the Portland Pattern Repository is so close to troping wikis that you can basically replace the word "software" with "entertainment" in the purpose and it makes sense.
I also reject the idea that software can automatically be used for free if it's open source.  I mean, look at Unix.  That's like the classic example.  It took some 30 years until we got a FOSS alternative to Unix and you're not really concerned *that* far down the road, are you?  There are legitimate reasons to keep your application closed-source, but at this point it seems like a matter of taste.

Since our wiki is built upon a real DB, we can rate all the pages,
whether they are good clean family fun or hentai guro. Tropers can use
their settings to decide which pages they want to see. If you don't want
to see the real bad stuff, you'll get a 404 when you go there by
mistake. Oh, and people will have to make an account before they can
read up about bad stuff. Good enough?

Actually, that's elegant.  I worry about the edit wars on what acceptable concepts are, but this is at the core a good idea.  Also don't send 404 on a valid request for a resource that exists -- 403 would be a better status but 200 would also work with a warning page.
Most of the ideas I'm coming up with involve forcing Mediawiki to do things that it really wasn't designed to do.  I have the core of an idea/implementation to enable Pmwiki traces (a.k.a TVT index footers) into a Mediawiki footer.  I'm not really sure if people even want that feature, but it's something that could be toggled.  I think that if I was truly clever with Javascript, I might be able to come up with something that mimics your example sorting thing.  No new ideas, I'm afraid.
-- ∇×V