The article mentions how setting up SourceHut from scratch is a complex undertaking with hundreds of small tasks. Some of those are understandable, especially given the amount of data they surely handle, but setting up a complete environment and restoring from a backup should be a simple and mostly automated procedure, not a gargantuan undertaking that needs to involve the entire team several days. There are always difficulties when production is down, and you're trying to restore a full system while undergoing a DDoS attack, I get that, but the reason we have modern cloud providers and tooling is to make creating new environments as painless as possible. It seems foolish not to take advantage of that.
I'm not a fan of Kubernetes either, but it's good that they're experimenting with it. Hopefully it leads to quicker deployments if this happens again.