If the rpm database becomes corrupted somehow, such that beah cannot install task RPMs, the task result in Beaker appears as "Completed New" which is not really sensible. The harness should at least report the task as "Completed Failed" or better yet, simply abort the whole recipe.
The task should have been aborted, which used to work fine. What does the job-results show?
*** Bug 877869 has been marked as a duplicate of this bug. ***
As the task logs of the original bug report show, although the abort event was generated by beah, it wasn't handled correctly. Discussing with Marian Csontos, we inferred that the problem was possibly in the beaker backend - beakerlc.py. If the async_proc (on end event) sets the task to finished before proc_evt_abort gets to it, it will return without aborting the task. This is what possibly happened here. Hence, checking for the completed status of the task should prevent this from happening. I couldn't ofcourse reproduce the problem when I run the job for a couple of times. On Gerrit: http://gerrit.beaker-project.org/#/c/1563/
We were never able to reproduce this - the change was made by analysis. Testing indicates that this change at least did not introduce any regressions.
Beaker 0.11.0 has been released.