Oh, I was more asking about the (potential) advantage of not using Treesitter.

I’ve used the incremental parsing from Treesitter in Neovim (though not on any absolutely massive codebases) and found it to be great (when it worked). But I suspect this is immaturity in the ecosystem (one of the Neovim implementation, the parser definition or Treesitter itself).

As I say, I haven't used Treesitter so I can't say, but in my experience most options do not perform well on huge codebases (> 5 million loc)