High: crmd: Bug BNC:467995 - Delay marking STONITH operations complete until STONITH tells us so
If STONITHd takes a long time to report the operation complete, the node
may have already restarted and rejoined the cluster. This means that when
tengine_stonith_callback() erases the node_state entry, its erasing the
correct details instead of the pre-stonith values.

Mercurial revision: 8969aac54cf38e2c8c1c63e112458b99b466df75