Anyway, forbidding pages from loading secondary content would break millions of sites, including the most visited sites in the world. That would be equivalent to completely redesigning HTML/JS.
"Only allow play of audio in response to user action."
Okay, cool, so there's a giant 'click' event handler on top of the whole page. When you click it I'm going to play a 250ms long sample of silence embedded as a data:// URL into the audio or video element.
Now I control the player and can do whatever I want.