Most platform teams don't over-architect because they're careless. They over-architect because every single decision was reasonable. Here's the one I see most. A team adopts a service mesh in year one. They'd read how a hyperscaler routes traffic. They copied the shape of it. The intent was good engineering. Two years later they're running a control plane, a data plane, and thousands of sidecars. To route traffic between a few dozen services. The mesh wasn't wrong. It was early. By about...