If it's BGP, wouldn't it be known by outside parties, since it is logged and viewed by, for example, Cisco and Cloudflare? (Also from my personal experience on the outage, I can reach the video server but it timeouts with a proper error code, so it signals that it is beyond external-facing networking. So either bad internal networking or more likely a bad config change or stack deployment.)
When I was at google there were a couple of times that "big services" went down. Every time the service was actually up, some change or another broke the connection from the gateway to the internal system. It's probably that. If it was just a bad push of a single component, usually you can revert it. If you push a bad gateway config, maybe the gateways are crashing so much you can't roll back.
