This is great! I read through the paper and watched the video, I want to try implementing this in wgpu. For pixel-aligned 2D text, I'm wondering if a hybrid approach would be good, where you use traditional glyph atlases for smaller font sizes with the right hinting applied, and then use this scanline sweep approach for larger sizes or where you know you'll be animating the size or viewing it in 3D with perspective.
Speaking of font hinting, if you're rendering with this technique, couldn't you also used the hinted bezier curves if you know you're drawing at a particular size and particular location on the pixel grid?