Do not know what the test plan looks like for self-heal but I would like
to suggest that reboot tests be made for 1.4 and self-heal. The ability
to survive a spurious production server reboot and have all cluster
nodes recover fully is one test that would be critical. Whether
unplanned, like kernel oops and crashes, or planned, like server
maintenance or recovering a hung process, reboots happen all the time
and it will be critical for GlusterFS to fully survive reboots and heal
without manual intervention.