Essentially, rendering JPEGs directly as a GPU texture (since GPUs can already natively use 420 encodings for textures). His point is that this would improve performance, very slightly improve quality, and reduce power usage as there's no colour space conversion.
4:1:1 is half of RGB but I've never seen anyone ever use it. Pretty much everything in use today (at the consumer level) uses 4:2:0. It's good enough for today's standards (AV1 Main profile only supports 4:0:0 and 4:2:0).
Assuming infinite precision, sure.
Interestingly however, there's a closely related color space called YCoCg, that actually does achieve lossless coversion to/from RGB (this lossless variant is often referred to as YCoCg-R, the R stands for "reversible"): https://en.wikipedia.org/wiki/YCoCg#The_lifting-based_YCoCg-...
YUV to RGB conversion (and vice-versa) requires use of conversion coefficients (BT.470, BT.709, etc): https://en.wikipedia.org/wiki/YUV
Different apps/algorithms can choose their own coefficients, so you get a slightly different RGB colors. If you converted RGB back to YUV with a different set of coefficients, you would get a different result.
The difference between 4:1:1 and 4:2:0 is that in 4:1:1 the chroma covers a 1x4 line of pixels whereas in 4:2:0 the chroma covers a 2x2 square of pixels.
YUV format is lossy. The one that is twice as bandwidth-efficient as RGB records four times less color difference than full bandwidth YUV or RGB. Add to that that RGB often used with alpha channel and for half the bandwidth you will get twice the artifacts.
The subsampling scheme that is half of RGB is called 4:1:1: https://en.wikipedia.org/wiki/Chroma_subsampling#4:1:1
It is not broadcast quality. 4:2:0 is also not good for today's standards.