kern.dfstarv.report: started_ms=1791334958871 victim_ran=1 victim_ran_ms=1791334959401 churn_count=151 waiter_done=0 waiter_woke_ms=0 now_ms=1791334961881 starve_ms=-1
    -1     -1 ??  B0       0:00.00 (dfwait)
    -1     -1 ??  B5       0:00.00 (dfhz)
=== T+9s (victim done since T+0.5s):
kern.dfstarv.report: started_ms=1791334958871 victim_ran=1 victim_ran_ms=1791334959401 churn_count=1000 waiter_done=0 waiter_woke_ms=0 now_ms=1791334978862 starve_ms=-1
=== stop churn:
 -> === T+10s:
kern.dfstarv.report: started_ms=1791334958871 victim_ran=1 victim_ran_ms=1791334959401 churn_count=1001 waiter_done=1 waiter_woke_ms=1791334978882 now_ms=1791334979862 starve_ms=19481

# ---------------------------------------------------------------------------
# run 2 (fresh kldload, 20s window) — run.2.log
kern.dfstarv.report: started_ms=1791334996922 victim_ran=1 victim_ran_ms=1791334997462 churn_count=1001 waiter_done=0 waiter_woke_ms=0 now_ms=1791335016942 starve_ms=-1
kern.dfstarv.report: started_ms=1791334996922 victim_ran=1 victim_ran_ms=1791334997462 churn_count=1003 waiter_done=1 waiter_woke_ms=1791335016982 now_ms=1791335017942 starve_ms=19520
UNLOADED-CLEAN

# Reading: the victim message completed at T+0.53s (victim_ran=1,
# victim_ran_ms). At T+20s the waiter was still asleep (waiter_done=0)
# although every queue message cost one 10ms tick (churn_count == elapsed
# ticks; no stall anywhere). 20ms after stop=1 (queue drain), the waiter
# woke: starve_ms = 19481 / 19520 ms of uninterruptible oversleep.