Early in my time on the Kubernetes team, a customer proposed something that was both brilliant and beyond what we were ready for: a global footprint of clusters, one per region, with a synchronized set of jobs. They were running a low-latency application and wanted to coordinate workloads across continents.