INEFFICIENT TOKEN USE AND WORKFLOW : Feature Request - When Sub-Agents are interrupted, allow the same sub-agents to continue working

When a sub-agent is conducting tasks and the user interrupts the workflow because something is wrong or needs to be changed, the process returns control to the main agent, which then either responds without the context that the sub-agent already had (forcing it to try to require the same context - which wastes tokens) or it takes a completely different path, potentially leaving work half finished that the sub-agent had started on.

It would be better if a task interruption of a sub-agent resulted in the same sub-agent handling the interruption and user feedback, since that's probably where the feedback is best directed, and then the change in direction could be implemented but eh sub-agent rather than relying on the process re-continuing by the main agent in the same way that it had been going.

The current behavior is very frustrating and I am going to disable sub-agents until this can get sorted. All too frequently, I need to see the main agent relearn and re-investigate (or spin up a new version of the same sub-agent to do the same) based on the user feedback after the interruption.