DF-4025 / run.log
---dmesg--- rereplica: loaded (hz=100) rereplica: armed (re_timer shape, 1-tick period) rereplica: re_tick parked on our serializer (fired=720 ran=719) - entering callout_stop() rereplica: IF THIS IS THE LAST rereplica LINE, AB-BA DEADLOCK IS PROVEN ---ps--- 1120 D5 costp sysctl kern.rereplica.stop=1 ---stop4025.out--- kern.rereplica.stop: 0RUN_DONE SECOND_SAMPLE_DONE 1120 D5 costp sysctl kern.rereplica.stop=1 --- post-test --- The wedged guest did NOT power off cleanly via ssh (callout starvation impact evidence): 'vm.sh down' timed out waiting for poweroff; QEMU force-killed and reverted to the 'with-src' snapshot; guest verified back up afterwards. Wedge observed for >20s across two samples before reset (permanent, not transient).