In this case, if it uses rrweb, it just stores all the HTML and all DOM mutations that happen. This is a pretty accurate method (even though still doesn't record all events, for example text selection, and if you watch an mp4 video for example, it won't save the entire video, only the URL to it), but it uses a lot of data.

In my case, I created a more simple solution for userTrack.net, where I just store the URL, record all events that happen and replay them by loading the iframe at the same URL and re-triggering all the events. It's not as accurate as storing HTML and mutations, but it uses a lot less data.