I’m skeptical of this claim. Signal doesn’t seem like it’d be very compute-heavy, doesn’t seem like text and voice would be very network-heavy, and I don’t think video is used very much. If us-east-1 going down took out most their services, it doesn’t seem like they’re leveraging AWSs multi-region features very well anyways. It wouldn’t be too hard to just rent or co-locate hardware in multiple non-hyperscalar data-centers around the world, and run a multi-zone, highly available k8s cluster. Would probably be cheaper and more robust too. I don’t have experience with multi-zone k8s, but I was the sole person responsible for deploying and maintaining a highly-available single-datacenter k8s cluster on rented hardware, and it wasn’t even my primary job (was a full-stack engineer and team-lead), If I could do it, I don’t think they’d need to try to hire world’s top experts or anything. Coincidentally, the provider was UpCloud, which is a European company, and in 8 years of using them, I don’t recall seeing a single node we had become unresponsive for more than 5 minutes, and I’m not even sure those times were on UpCloud’s end.