An intriguing hypothesis, is there any data to support this theory ?
More of the stuff that will increase a user’s watch time; subtle difference. I’d be surprised if they haven’t also experimented with adjusting for overall ad revenue; some specific ads pay more than others.