Well, for once, it does offer a certain fexibility - a custom renderer for a game will always perform better than an all-purpose renderer written for an all-purpose game engine.
Besides, there's several reasons why I personally would say it's worth it.
First of all, this specific renderer was written for a practical course at University (I'm a computer science student). As the title in the window suggests, we have to create a small game aimed to be played with the Oculus Rift. Using another engine wasn't any option anyways for this project, as it's basically the point of the course to write an own engine. I'm mainly responsible for the graphics programming, together with another student who has implemented the very simple shadowmapping seen in the screenshots as well as skeletal animations and the import/export pipeline for the 3d assets. Lately I've refactored the whole rendering pipeline because it was rather inflexible and didn't support gamma correct rendering and HDR, which it does now. And boy, did that feel good. What brings me to the second point, there's people who only make games using the tools they can find, and that's fine, but there's also those who actually create these tools (and sometimes make games with them theirselves), and if I had to choose one path I'd choose the latter. I just like doing things from a certain level of scratch, so to speak.
There's a high level of satisfaction when a game you programmed is fun, but that's nothing compared to the level of satisfaction you feel when you know that game runs on your own engine.

So to sum it up, I just like doing it, it gives me kicks that simple game programming doesn't, and it's simply cool as fuck to have your own engine.