Looking at the source code, it seems like I should do:
data_processing_task._artifact_manager.flush() to make sure to have the latest version of artifacts in the task, right?