DF-2630 fix_run.log — kernel #1 (fix.diff applied; battery identical for
kernels #1 of 02:23:41 / 02:52:56 / 05:10:55 UTC 2026 — the 2630 hunks
are unchanged across those builds).  Decisive lines verbatim from the
sessions (root_timeout demo value 5 s).

vfs.usermount: 0
=== A. sanity: stock single-device image mounts/ls/umount fine ===
SANITY_MOUNT_RC=0
SANITY_LS_RC=0
SANITY_UMOUNT_RC=0
=== B. control image (pfs_nmasters=1): no wedge ===
CONTROL_MOUNT_RC=0
CONTROL_LS_RC=0
CONTROL_UMOUNT_RC=0
=== C. wedge image (pfs_nmasters=2) with root_timeout=5 ===
vfs.hammer2.root_timeout: 60 -> 5
WEDGE_MOUNT_RC=0
1787967678                      <- ls #1 start (epoch s)
ls: /mnt/h2: Input/output error
WEDGE_LS1_RC=1
1787967683                      <- ls #1 end: exactly the 5 s bound
--- ls #2: sticky-fail immediate ---
1787967683
ls: /mnt/h2: Input/output error
WEDGE_LS2_RC=1
1787967683                      <- same second: instant sticky fail
--- census: any h2root/vfs_busy waiters? ---
CENSUS_DONE                     (zero waiters)
--- dmesg: quorum failure message ---
hammer2 (/dev/vn0@testvol): root quorum cannot be reached (nmasters=2
nchains=1), failing VFS_ROOT after 5s
--- other hammer2 mounts still possible (system healthy)? ---
HEALTH_MOUNT_RC=0               (fresh othervol image on vn1)
HEALTH_LS_RC=0
--- can the wedged mount be removed? ---
WEDGE_UMOUNTF_RC=0              (umount -f REMOVES it cleanly)
--- final state ---
VALIDATION_C_DONE               (zero h2root/vfs_busy/ncplk waiters)

Post-run stability on the guard-carrying kernels:
  T+60S_ALIVE / T+125S_ALIVE / T+160S_ALIVE / T+280S_ALIVE
  testvol_xop=1 (grep only — zero leftover workers)
  dmesg Fatal count = 0