Scraping is an arms race, of course. A simple but often successful way to fight scraping is, for instance, changing the names of classes routinely. This affects scrapers more than users, because a user doesn't see those class names, while a scraper relies on them.

Is anyone scraping using only (or mostly) computer vision on the rendered browser screen, and simulating mouse clicks and key presses?

It seems like anti-scraping measures to defeat an CV-based approach would be more intrusive to the user and thus they would be used less often.

[0] https://github.com/jimbobewenhall/OpenCV-website-scraper [1] https://incolumitas.com/2021/05/20/avoid-puppeteer-and-playwright-for-scraping/