Author here.
Thank you for visiting and for writing about it. It is indeed a bit clunky, and the data download is indeed dumb, and these are both in my todo.txt. Now that I've done a bit of prototyping, I plan to rewrite the data output streams so that it's easier to write pages which produce, all at once, html, rss, zip, and maybe something else, too.
The forum data is not just merely downloadable. The neat thing about using PGP for auth is that you can clone the forum from this data, rebuild it, and the same identities can be used on both forums, with full continuity of user accounts. The accounts themselves, the private keys, remain with the client.
Yes, you could potentially transfer the messages between different servers, as well as have each server tag them too.
TeX support is not planned at this time, but I don't see why that wouldn't be possible. What I do want to ensure is that when a user types something in, it is rendered how they meant it, not how the programmer thought it should be rendered based on some bullshit rules, e.g. Markdown or <ul><li>. I want to eliminate the possibility of a user typing 1. 3. 5. and seeing 1. 2. 3. in the output, or typing http://wikipedia/blah_(foobar) and having the last paren drop off, that's a bad user experience.
I have indeed optimized for all the browsers I've had access to, which is currently Windows-centric, like Netscape, Mosaic, IE, Opera, OffByOne, and Safari, but also includes iOS 7 and 8, recent Mac Safaris, occasional Chrome, Android Chrome including older ones, Links, w3m, Dillo, NetSurf, Lynx, Seamonkey, Konqueror, Midori, and qutebrowser, my favorite browser at the moment. Most have been tested with JS and without, and there is also an Light/Accessibility mode, which strips all but the most essential tags from the pages. It has recentl been tested by a screen-reader user, but I have yet to delve into that myself.
I came here googling my own URL, and I really like this forum, and happy to answer any questions.