I just watched https://www.youtube.com/watch?v=Q4dYwEyjZcY this video about the early HTML standardization process, and it seems to explain all the ills of HTML.

So indeed, there is a huge optimization opportunity of having a stricter error model.

Also, I'm really wondering how much battery could be saved when surfing such pages.

Also I'm sure there is a lot of potential going in the pre-parsed document model. But that's a next level kind of engineering I guess.