Once you're streaming something live and interactively, if nothing else, you're getting boned by the limits of physics: for example, a connection from a datacenter in California has a (current) ~150ms minimum round-trip latency to Europe. That's bordering on a human-perceptible lag, and that's the best case. As I said before, most providers don't give you anywhere near the best case. So you spend a ton of money setting up datacenter and peering arrangements and low-latency interconnects, but you're still in a tough place if your broadcaster is connected to a craptastic residential ISP in New York, and your viewer is in Hong Kong.
Maybe YouTube can do better simply because they're YouTube. But I doubt it. Everything YouTube has done has been to optimize the case of static video delivery, which is a different (and much easier) game. Doing live video at scale requires a different architecture.
Even during soccer games, where there's often 20+ seconds between the radio & cable transmissions, I've never seen it being a problem. People just choose turn off the radio.
As for latency, Twitch -- the current de-facto -- already has multiple seconds of latency added on top of the encoder and decoder delays. Being aggressively realtime and uber interactive with your stream is not really a reality yet. You can right-click a Twitch stream to see these kind of stats; for example, a Game's Done Quick Stream: http://i.imgur.com/dcPoRui.png
YouTube promises two (well one actually) change that I welcome. It will allow one URL which I presume will go to the live broadcast and it promises that you won't have to create an event every time you start broadcasting.
With these two changes, I am ready to switch to YouTube at any time.