We just spent months at my workplace working on a system to reliably define and configure a set of parallel silo'ed integrated datastores, services, and network

We just spent months at my workplace working on a system to reliably define and configure a set of parallel silo'ed integrated datastores, services, and network stack within Kubernetes/ISTIO (and AWS), and to reliably upgrade new software revisions within those silos and to account for the changing "shape" of the configuration/content in these silos. It's repeatable and safe now, but it took a lot of effort.