It's a hard problem to figure out what's readable text on a page, and what isn't. Even Google has a hard time figuring that out. OCR works very well with screenshots, and is purely computation time. But the real reason is generally just having timestamps, urls, and screenshots is good enough. I usually remember about when it was, and some words in the url, and don't need the heavyweight text search setup.
Trying to parse the SPAs of today is just painful. Simpler to just render the page screenshot and OCR! Guaranteed to only index text that actually matters
