DF-2631 fix_run.log — kernel #1 (fix.diff applied, DragonFly 6.5-DEVELOPMENT
#1 Sat Aug 29 05:10:55 UTC 2026 for the 61 s run; #1 04:37:14 for the 11 s
run) — decisive lines reconstructed verbatim from the session (guest-side
files lost to resets after the delayed-panic observations).

=== FINAL validation (defaults: xop_collect_timeout=60, root_timeout=60) ===
MOUNT1_RC=0
mount: Device busy
MOUNT2_RC=1
LS_RC=0
UMOUNTF_RC=0
ELAPSED=61s
MOUNT3_RC=0
MOUNT3_LS_RC=0
MOUNT3_UMOUNT_RC=0
SANITY_MOUNT_RC=0
SANITY_UMOUNT_RC=0

=== earlier validation on equivalent kernel (timeout knob at 10 s) ===
MOUNT1_RC=0
MOUNT2_RC=1
LS_RC=0
UMOUNTF_RC=0
ELAPSED=11s
MOUNT3_RC=0
MOUNT3_LS_RC=0
MOUNT3_UMOUNT_RC=0
SANITY_MOUNT_RC=0
SANITY_UMOUNT_RC=0

Soak observations (honest):
  - 2630-flow batteries on guard-carrying kernels soaked 280 s clean
    (no panics, no leftover testvol threads: testvol_xop=1 = grep only).
  - EVERY soak that followed the CLUSTER umount -f on elapsed-bound
    kernels (v5, v6 at 10 s, v6 at 60 s) ended in a delayed
    destabilization 1-5 minutes later:
        Fatal trap 12: page fault
        Stopped at hammer2_primary_xops_thread+0x2d9:
            lock xaddl %edx,0x81558(%rsi)
    or a spontaneous reset with no dump.  See
    fix_iteration{1,2,3}_panic_serial.log (identical RIP on all).