What they're doing is not bad: -13 gives plenty of room to get a sound in, and it naturally fits with a 'hey I don't have to turn this track down anymore' discovery period. They can't get away with something as extreme as -23.
I believe it is better if mastering engineer carefully clips, limits the signal to make it loud and as clean as possible, than an end user clipping his internal soundcard dac by 7db for example. But what happens if a modern dance song is mastered dynamic, and there is no those 7db of headroom but user still pushes it with some gain booster ? I'm afraid that end results can be pretty awful. And if you have, for example, a -5Lufs song, for example, youtube streams it -7db quieter, in theory the user will have 7-ish db more to push without audible artifacts. I'm afraid that some kinda of gain boosters will appear that will adding gain on top of the the audio people have on their soundcard.
Plus one more thing I wanted to ask you guys, I already seen few people (laptop users) complained how they were noticing that youtube was "quiet" and that was driving them nuts. I'm afraid that this may result in another loudness war, in another words, making a song appear louder on youtube, especially if we take the lowest consumers in consideration that doesn't have low end extension in their devices (laptops, phones, the cheapest earbuds.), that would result with bunch of mid pushed, and bassless songs just to appear louder, I imagine. While I was listening to the music on youtube yesterday I realize that, unfortunately, the songs that are mixed/mastered to be loud are still kinda louder then dynamic ones, psychoacoustically, like more mid forward, more harmonics, less bass to take the range or headroom or whatever.