Only members with permanent status recording can become failed. A member in the failed state is one whose associated task, job step task, address space, or system terminated before the member was explicitly deactivated by invoking an XCF service. When a member is in the failed state, other members can infer that the member did not have an opportunity to clean up its own resources, and another member should take recovery action.

For address space associated members, however, I/O is purged as part of address termination cleanup before the group user routines of the surviving members receive control to inform those members of the failure.