I have flows directory in my project with a parent flow in file parent.py, which calls a child flow in child.py. The child flow consists of multiple mapped tasks. When I run the child flow directly, or if I import child.py into parent.py and invoke the child flow directly, everything works as expected. However, if I instead use flow.from_source and pull child.py from a Git repository, the subflow that runs only executes the first task and the final task in the sequence , skipping everything in between. The only thing I can think of that might be a factor is that I have a class definition in child.py that's outside the scope of the flow & task definitions - is it possible that this is being ignored when I invoke the flow from source? I'm also having problems using blocks in child flows. In the same example, I pull a JSON secret in the child flow and pass the results to a task. Once again, running this flow directly doesn't have a problem, but running it as a subflow fails. What's bizarre about all of this is that I have other flows that are being invoked from source and they seem to work just fine, and they also use blocks. I'm at a loss. Any idea why I would experience different behavior when I run flows from source rather than locally? The whole reason I'm trying to do this is because I'm using ACI/serverless push, and I want to avoid building a new image each time my code changes slightly. I want to have a standard base image that kicks off a parent flow, which in turn pulls from GitHub and executes a sequence of subflows.