Dunno what actually happened, but one point of vulnerability that could bring down an entire service like that could be their internal DNS management SW. One IT staff screw-up and the SW can't match incoming requests to the location of the content in their internal network. This happened a few months ago to a major hospitality and entertainment company and customers couldn't access anything for a couple of hours.
As for DUpipo's Slav-Paranoia ...
This is my suspicious as well.
The root name server for the zone controls the zone serial number. a serial number changes forces out an update all the way down the chain of propagation to name servers all over the globe.
If this were a remote hack or something, we would have seen sporadic or localized outages spread across time. This appeared to be almost instantaneous, absolute, and complete. That points to a potential issue on the root name server for that zone... be it accidental, or sabotage. No single device or link could possible cause this sort of massive outage on this scale.
That is UNLESS youtube has some similar propagation system for rolling out new code. But I would think they would test that regionally first rather than risk a global slip-up.
Then again, we are talking about people... and they screw up all the time, so who knows I suppose.
As much as I despise google, I have to give them credit. I only WISH I had the uptime numbers they can brag about.