Really looking forward to the the stereo rendering of non-WebGL content, HTML+CSS, which the Mozilla team is working on. More challenging path to figuring out but will open a lot of possibilities.
In particular, we have some ideas for preserving depth information between CSS and WebGL. Imagine, Z Fighting between something rendered in WebGL and something else positioned with CSS. Or I think a more compelling use case: correctly positioning an Iframe in a WebGL scene.
It's something we're looking into, but will take a significant amount of investment in platform (two very different parts of the rendering engine suddenly have to communicate).
For now, it's very low hanging fruits for content creators or engine developers to work with WebGL. It also doesn't hurt that it's much more performant.
Yes ideally they'd be able to be composited together at the depth buffer level. But even separate I can see a ton of devs making some very cool more UI-centric applications just with CSS mode. Can't wait to see the webdev community understand what can be done with VR and the webstack in 2016.
It's something we're looking into, but will take a significant amount of investment in platform (two very different parts of the rendering engine suddenly have to communicate).
For now, it's very low hanging fruits for content creators or engine developers to work with WebGL. It also doesn't hurt that it's much more performant.
On the other hand, it hurts accessibility.