If you have any of those pages open, though, you’re still a part of the experiment and of course will see the results as expected. Just as I assume if you open a malicious website (via a popup for example) using this method, as long as that window/tab is open, it has access to and can report your real-time browsing information about at least 30 sites. Which isn’t a significant number, to be fair.
Given that many security researcher’s job is to make exploits sound as scary as possible without being so hyperbolic as to become a laughingstock among other researchers, the fact that the security researcher “suspect this number to be significantly higher in real-world scenarios” without actually checking and providing further info on some of their suspicions about databases on sub pages, those specific user actions or authenticated parts of pages, you can almost guarantee that the exploit provides incomplete information regarding connections to 30 sites. If the number of sites on the web were 40, perhaps, this would be more concerning.