df4276: replica armed
df4276: event thread tail complete; final wakeup() delivered into the void (detacher not yet asleep)
df4276: detach COMPLETED -- no lost wakeup
df4276: detach_td=0xfffff800905e0700 wchan=0xffffffff81313670 wmesg="nanslp"
df4276: RESULT: no lost wakeup

# Interpretation: the forced lost-wakeup interleaving does NOT hang the
# detacher. The tsleep(event_thread) returns because lwkt_switch_return()
# unconditionally wakes sleepers on an exiting thread's ident
# (sys/kern/lwkt_thread.c:884-899). The 'cbbun' lost-wakeup hang axis of the
# original hypothesis is DISPROVEN; the surviving defect is the
# delete-children-before-thread-stop ordering (see VERDICT.md).