I suspect this is due to how we synchronized hooks / juju-run. The fix for bug 1486712 changed the locking mechanism used, and this may fix the problem.
Previously, a killed process wouldn't relinquish the lock. But the new implementation shouldn't have that problem.
Once 1.25.6 is released (hopefully soon), could you retest to see if the problem persists?
I suspect this is due to how we synchronized hooks / juju-run. The fix for bug 1486712 changed the locking mechanism used, and this may fix the problem.
Previously, a killed process wouldn't relinquish the lock. But the new implementation shouldn't have that problem.
Once 1.25.6 is released (hopefully soon), could you retest to see if the problem persists?