The key is that they're not linking remote content – they copied the content of the tweets into their websites' code, and then used Twitter's JS to render it. I.e., it used to be a purely stylistic change.

All of a sudden, this very inert code has changed, without warning, to actually edit any websites using it to delete the text that they put there. I think many would consider this malware-adjacent, if it wasn't from such a large company.