This would mirror the behavior of the synchronous task group.
One could maybe want a task group that recovers from error but we don't have a need for it today and, since the asynchronous task group only has a single status it, the current implementation isn't well suited for that use case anyways.
Reporter: Weston Pace / @westonpace
Note: This issue was originally created as ARROW-14565. Please see the migration documentation for further details.