Summary

Removes a panic source in the conductor frontend by moving around the restart logic.

Background

Conductor::run_until_stopped contains a loop to allow restarting its backend business logic. This loop delegates to a Conductor::restart_or_shutdown method to execute the restart of the backend, and then immediately reenters the loop no matter a restart happened or not.

While refactoring the inner conductor business logic (which lead to slightly different return values in turn affecting the blackbox test assertions) it was found that behavior can lead to the conductor backend panicking because it's being polled again even after running to completion.

Changes

• Rename Conductor::shutdown_or_restart to Conductor::restart_or_shutdown to match the enum it's branching on.
• Execute the restart inside the frontend loop to not reentering it, avoiding to poll the already completed backend task.

Testing

This can't really be tested. But it does fix the panic in the refactor PR.

Changelogs

Changelogs updated.