

They take advantage of viewer federation a lot, webrtc is used, so all the simulated browsers are sending the video to each other rather than hitting their server. So their setup is really just a part of the p2p swarm, like a single client in a bitorrent network. Doesn’t use anything fancy above that.
Honestly, it’s only a setup that’s gonna get far if you serve very few videos, and the P2P client rate is high. Their “real world” assumption of 50% of all clients being p2p enabled is way too high, and they couldn’t limit p2p bandwidth so all the clients were sending data to each other at lan speeds.
It’s interesting, but it needs to actually leave the simulation and enter real world load to know how things shake out.

Azure products ask you for your identity and signin a lot. Honestly, I’m asked to log in again at least once every 24 hours. That’s assuming I don’t traverse some sort of service wall where I’m now in a different system after clicking a link.
I do cloud engineering for a living, and I would probably fall for at least some phishing things around Azure, specifically because azure identity management is so obtuse and constantly asking for things.
It’s absolutely on the system that Microsoft designed , and the practices they encourage, and the mitagations that apparently don’t exist.