This should be fairly easy to reproduce with a simple stack (only one resource required). - set a pre-update hook on a resource - start a stack update - wait for the workflow to reach the hook and pause - restart heat-engine while heat is still waiting for the hook - do another stack update, removing the pre-update hook from the environment (so that this update should not stop for the hook) - ensure that the update completes Previously, the last update would stop at the hook (even though it wasn't requested in the environment), although you couldn't tell that because nothing appeared in the events or 'heat hook-poll'.