I'm not doing it from scratch, really. I have a code base built up from a few years of work. In it, I have everything from A* search algorithms to various library implementations (FBXSDK, libpng, OpenAL, etc.)
In addition, I've had the opportunity to work on some other cool stuff like glyph streaming (for localization) and vector shape streaming for things like menus. I don't like licensing middleware. If you want something done correctly, you must do it yourself.
Pipelines for content are really nontrivial. I'm already familiar with the FBX format, which almost every modeling package uses. It's what is used to get skeletal meshes into UDK (and others like Unity.) Animation is only slightly more involved than figuring out where the vertices are supposed to be at what time, and blending between the different animations.
The reason why this is reasonable is because we wouldn't lack focus or direction. Xreal is, from what I can tell, ioQ3 with a nice renderer. That's all it is, though. It's not a game; people cannot play it. If my engine becomes great enough, I'll look at turning it into middleware. That's currently not my goal. My goal is to produce something I can be proud of. The only way to achieve this in a pure way is by taking the hard route.
There are other numerous reasons why I would want to use my own engine instead of ET. For one, licensing reasons. If we all decide to turn this into an ad-powered f2p game with online stats tracking and ranked matchmaking (in addition to dedicated servers) then there would be no legal hurdles. It's my code. The only libraries I use are the MIT/zlib/BSD-licensed ones (free for commercial use.)
PLEASE don't be pessimistic, coroner

we've already got a wealth of content to use for testing purposes. This in itself greatly speeds up the dev process. Also, because the engine is still in its somewhat-early stages, its development can be easily influenced.
I'd like to encourage you to talk with me on Skype, MSN, or IRC sometime. I think I can convince you that this is a real possibility
