So if a flow calls another flow function directly, the subflow function code runs in the same process+infra as the parent flow function; but it sounds like a flow can instead launch a deployment, in which case that subflow gets launched on new infra and is monitored remotely via the API? Either way, the flow runs for all subflows will still appear in the UI as shown in the slides?

The first part is all correct. It may look slightly different in the UI in the two cases. I’m trying to remember if the flows show as linked by arrows in the run_deployment case. I don’t believe they do, but I believe there is an issue for that in the repo.