This prevents *very* long waits while attempting to destroy channels
after a fault has occurred.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>