# DF-2645 run.log (summary of both stress runs; guest files were on the
# pre-reset guest, excerpts preserved from the captured sessions)

Run 1 (mode 1, 12 min, pre-collapsed topology):
```
created 6000 files
tree built: 6000 entries
deleted 4200/6000   (70%)
survivors: 1800
starting stress (mode 1, 12 min)...
w0..w3: ~34M-47M appends each (total >130M), churn child active
stress done without panic
    1800
UMOUNT_OK
```
dmesg afterwards: no LOST CHILD / debug repchange / excessive loops.

Run 2 (mode 1, 10 min, continuous key-space-wide churn):
```
starting stress (mode 1, 10 min)...
w0: 47579136 appends ... churn0/churn1 progress lines
(teardown stalled)
```
dmesg afterwards: thousands of
  hammer2_flush: 000000001xxxxxxx.0{1,2} error=00000020   (ENOSPC)
  hammer2: unable to fsync inode ...
  Warning: vfsync skipped 2 dirty bufs in pass2!
`sync` PID stuck in h2coll state (14+ min CPU) — ENOSPC flush-retry
storm on the full 512MB volume (DF-2633 family; NOT claimed for this
finding). No maintenance markers.