I'm most interested in launching an interactive environment from a failed child task using the task's container with the input data for that task available for debugging purposes.

For debugging purposes, I don't think that launching at the Run level helps as much b/c what you'll likely need is access to intermediate files produced at some point in the run.

Updating to include launching from a task.

Neon carrot Goldfish Thanks for your feature request. We are currently implementing a "Launch Data Studio" from the Pipeline Run details page at the parent Run-level (which includes the same image with the input data copied into the environment), however this isn't at the child Task-level. Does having this functionality at the Run-level satisfy your request?