Unfortunately that destroys accessibility. If you want to support accessibility, then you need a more complicated rendering target than something like a canvas.
It sucks but this is the way we're heading. I'm not sure web devs will even be writing HTML in 5 years beyond adding a canvas tag.
In fact, the approach would be more general, since a lot of text is already embedded in images on a lot of websites (think also about ads).
The simplest example: how do you indicate a graphic or piece of text is a button or link? How do I set the alternate text if it’s a graphical button.
Besides, nobody says we can't have a protocol for some additional accessibility information.
1. No-one uses it
2. There is only 1 implementation, and it is never updated
And then we wonder where all the software bloat comes from.
Not to mention that there's no such thing as "Easily run OCR", considering how many weird fonts there are, and that they could be rendered in such small size as to make it difficult even for humans to tell them apart.