I don't remember all of the details, so if you're curious you should watch the talk. :)
But IIRC the main points that differentiated this new effort from the old one were that (1) there was buy-in from the very top of the company, (2) they took a more incremental approach where the old one was more all-or-nothing, and (3) they tackled the edge cases and hardest repos first.
The talk includes some info about the failings of their first effort as well. It's worth a watch.
Yeah. I'm happy for them. But I'm also sad, because clouds are awful in terms of the way they handle state, and Garnix was paving the way towards something better.