That depends on how you’re running things. If you’re running a Kubernetes worker, it will run your workloads as Kubernetes jobs, and if the flow is paused, then it will not create those jobs, allowing your cluster to scale down. It requires some setup to get that autoscaling to work correctly though, particularly you will need to set resource requests/limits and use the cluster autoscaler.
If you’re running with Cloud Run or ECS, it will run workloads as Cloud Run Jobs or ECS Tasks, and if the flow is paused, then it won’t create those. So you should only need to pay for running the worker, which should be relatively inexpensive, since it’s pretty lightweight
If you use push work pools, then you don’t need to run an agent at all, so that allows full scale-to-zero while stuff is paused