Reason to get a false negative only one osd, when calling kill_daemons
$dir KILL osd, which means kill -9 shutdown() would not be called to
notify monitor so that it remain active+clean and osd up until monitor
got a mon_report timeout. Thus wait_for_clean lose it function and when
we call ceph pg scrub, we got an EAGAIN error and monitor will ignore
the scrub message because of osd not up