Follow-up to the fix in https://github.com/jenkinsci/jenkins/pull/4850 and JENKINS-61687. In some cases concurrent run deletion attempts may result in deletion not happening due to error while API callers will get a "success" result. Maybe it makes sense to actually lock requests and propagate errors to all callers if an issue happens. Without async API and timeouts such implementation would be error-prone on its own, so maybe the fix is acceptable as a simple patch. See the discussion after https://github.com/jenkinsci/jenkins/pull/4850#pullrequestreview-453722591 for details