DF-2631 run.log (stock kernel #0) — reconstructed verbatim from the live
verification session (guest-side file lost to vm.sh resets); every
decisive line was captured by the script's ps/grep stages on
Sat Aug 29 01:00-01:04 UTC 2026.

=== 0. prepare UNMODIFIED images: dd clones of one newfs image ===
MD5 (/root/poc/df2631/m1.img) = 956fc18a70f6689a8669e21d372b4eb9
MD5 (/root/poc/df2631/m2.img) = 956fc18a70f6689a8669e21d372b4eb9
=== 1. mount MASTER #1 ===
MOUNT1_RC=0
T1_h2xop=37          (36 workers + grep)
T1_h2nod=1           (0 sync threads + grep)
=== 2. same-clid merge via second device (EBUSY expected, chain merges) ===
mount: Device busy
MOUNT2_RC=1
T2_h2xop=109         (108 workers = 2 columns x 36 + ... + grep)
T2_h2nod=5           (2 h2nod-testvol + 2 h2nod-LOCAL + grep)
--- h2nod wchan census (merged cluster) ---
    0     -1      0   0 -12 -64      0      0 h2coll   B2     ??    0:00.00 (h2nod-testvol)
    0     -1      0   0 -12 -64      0      0 h2coll   B3     ??    0:00.00 (h2nod-testvol)
    0     -1      0   0 -12 -64      0      0 h2coll   B4     ??    0:00.00 (h2nod-LOCAL)
    0     -1      0   0 -12 -64      0      0 h2coll   B5     ??    0:00.00 (h2nod-LOCAL)
=== 3. ls works? (quorum 2/2 both masters present) ===
LS_RC=0
=== 4. umount -f  (expect: WEDGE in h2twait) ===
--- +3s: umount thread ---
   PID STAT   WCHAN    COMMAND
   932 D3     h2twait  umount -f /mnt/h2
--- +3s: h2nod census ---
    (same 4 h2coll lines as T2)
--- +8s: umount + h2nod + survivor counts ---
    0    932    930  24 198   0   4392    704 h2twait  D3     ??    0:00.02 umount -f /mnt/h2
    (h2nod unchanged, all h2coll)
SURV_h2xop=109
--- +63s: same census (post 60s tsleep timeouts) ---
    0    932    930  11 185   0   4392    704 h2twait  D3     ??    0:00.02 umount -f /mnt/h2
    (h2nod unchanged, all h2coll)
SURV63_h2xop=109
SURV63_h2xop_ALL=217
+ head -50 of h2coll/h2feed/h2twait/frozen census:
    0     -1      0   0 -12 -64      0      0 h2twait  B2     ??    0:00.00 (h2xop-testvol.32)
    0     -1      0   0 -12 -64      0      0 h2twait  B2     ??    0:00.00 (h2xop-testvol.26)
    0     -1      0   0 -12 -64      0      0 h2twait  B2     ??    0:00.00 (h2xop-testvol.20)
    0     -1      0   0 -12 -64      0      0 h2twait  B2     ??    0:00.00 (h2xop-testvol.14)
    0     -1      0   0 -12 -64      0      0 h2twait  B2     ??    0:00.00 (h2xop-testvol.08)
    0     -1      0   0 -12 -64      0      0 h2twait  B2     ??    0:00.00 (h2xop-testvol.02)
    ... (repeats per column; final histogram also shows h2idle workers
        of a second, lazily re-created cohort)
    0     -1      0   0 -12 -64      0      0 h2coll   B2     ??    0:00.00 (h2nod-testvol)
    0     -1      0   0 -12 -64      0      0 h2twait  B2     ??    0:00.00 (vbd0s1d)
=== 5. mount still present? third hammer2 mount blocks on mntlk ===
    0    972    971  21 195   0   6456    752 mntlk    D3     ??    0:00.01 mount_hammer2 -o ro /dev/vn2@testvol /mnt/h2d
   (MOUNT3_RC never written within 20 s; mount_hammer2 D-state persists)
=== TRIGGER COMPLETE — guest wedged ===

Post-run: `shutdown -r now` -> serial (shutdown_serial.log):
  Aug 29 01:03:41 init: some processes would not die; ps axl advised
  unmount of filesystem mounted from /dev/vn0@testvol failed (BUSY)
  unmount(...): forced umount of "(null)" - 1 namecache refs, 2 mount refs
Recovery: reboot (QEMU reset).  umount -f never completes on stock.