I can't substantiate my comment with academic sources, but here's my memory from the past decade: the people designing interface guidelines for mobile operating systems decided that carefully designing static visual assets for UIs was a waste of time (with similar technical complexity reasonings due to the number of different display resolutions/pixels per inch/aspect ratios)

Market context: Microsoft yearned to implement the Metro UI they'd been researching for a long time (Zune was its public appearance) — and Vista kind of soured people on Aero by osmosis; third-party Android applications always had consistency issues, so Google wanted to mandate something that would be generic enough to work (and be easily implemented) for all apps but also look distinctly specific to Android; Apple wanted to signal some shift to upscale high fashion (parallel to "everything is black glossy glass and aluminum" in its industrial design) and understood fashion-minded upscale as a mix of modernist typography with neon bright contemporary color schemes.

To make up for the boring visual aesthetic that often resulted, they recommended adding animations throughout apps. Most of these animations aren't very helpful, but they keep the user distracted for a moment.

I chose my words carefully: font compositing, not rendering. Font rendering still happens in software, but in order to composite it correctly into the surface requires DSB. I (hopefully) shouldn't have to explain the advantages of using hardware acceleration for app rendering in general.

High quality GPU font rendering is possible now.

http://sluglibrary.com/

I'm aware of Slug (and other GPU font rendering techniques). Slug, however, doesn't have subpixel AA (it doesn't make sense for it to, since its main goal is fitting in a 3D scene), and requires a moderately expensive (read: hard to do realtime) up-front bake process which converts the font into a pair of textures.