I'm doing some work on performance monitoring and have been investigating instances startup times. I see that GCP has the container startup_latencies metric. The description says: Distribution of time spent starting a new container instance in millis...