diff --git a/docs/re/f6-unit9-wrap-period-not-obtained.md b/docs/re/f6-unit9-wrap-period-not-obtained.md
new file mode 100644
index 00000000..52f44a02
--- /dev/null
+++ b/docs/re/f6-unit9-wrap-period-not-obtained.md
@@ -0,0 +1,49 @@
+# F6 unit 9 — the wrap-to-wrap period was NOT obtained: two harness failures, no new data
+
+**Status: ❔ not answered — the run did not produce data.** 2026-09-02.
+
+**Question:** what is the leaf's loop period measured wrap-to-wrap?
+**Look at:** frames between consecutive wraps. Pass = ≥2 wraps; fail = fewer.
+**Not covered:** everything else in F6.
+
+---
+
+## What happened
+
+Unit 8's rate (**0.514 units/frame**, leaf ≈ ½ × title) rests on **one wrap** —
+a start-to-wrap span, not a period. Hardening it needs a longer title capture with
+two or more wraps. Two attempts, both failed on the harness rather than the game:
+
+1. **First launch never started.** The `pgrep … && echo || { … }` guard I wrote
+   took the wrong branch — the output directory was never created and no process
+   ran, while a stale emulator from the previous iteration was still up. It looked
+   like a running capture for several minutes.
+2. **Second launch started, armed, pressed Ⓐ — and wrote no draw log.**
+   `canary.stdout` stayed at **0 bytes** and no `xenia_re_ui_draws_01.log` appeared,
+   despite the script printing `armed at 8s`. The likely cause is a race with the
+   orphaned emulator from failure 1, but I did not confirm it.
+
+## What stands, unchanged
+
+**Unit 8's two numbers are not affected** — they came from the intact `f6` capture,
+which is still on disk. The offset (+40 title units) has two independent supports;
+the rate (0.514×) still rests on a single wrap and **the port should not ship on
+it**, which is what I told them before this attempt.
+
+## 🔴 The lesson, since it is the second harness failure in two iterations
+
+**`armed at 8s` printed while nothing was being logged.** The script's arming step
+reports success on sending the keystroke, not on the logger responding — exactly
+the silent-failure shape that cost the first F1 probe a whole run, and which I
+"fixed" then by making the *window lookup* fatal. That fix was too narrow: the
+window was found, the key was sent, and the log still never appeared.
+
+**The check that would have caught both:** after arming, wait for
+`xenia_re_ui_draws_*.log` to exist and be non-empty, and abort loudly if it does
+not. A probe that cannot confirm its own instrument is recording is a probe whose
+negatives mean nothing — and I have now written that same class of bug twice.
+
+## Next
+
+Add that assertion to `title_sweep_probe.sh`, then re-run. The measurement itself
+is unchanged and cheap; only the harness needs the guard.
diff --git a/tools/re-capture/title_sweep_probe.sh b/tools/re-capture/title_sweep_probe.sh
index 7a26e3f4..35a3c8d2 100755
--- a/tools/re-capture/title_sweep_probe.sh
+++ b/tools/re-capture/title_sweep_probe.sh
@@ -17,7 +17,7 @@ OUT="${1:-/sylph-home/re/f6}"; mkdir -p "$OUT"; rm -f "$OUT"/xenia_re_ui_draws_*
 pad(){ printf '%s' "$1" > "$XENIA_PAD_FILE.tmp"; mv "$XENIA_PAD_FILE.tmp" "$XENIA_PAD_FILE"; }
 pad ""
 ( cd "$OUT" && nohup run-canary --log_ui_draws=true \
-    --ui_draw_capture_frames=2500 --ui_draw_capture_max=600000 \
+    --ui_draw_capture_frames=6000 --ui_draw_capture_max=1500000 \
     >"$OUT/canary.stdout" 2>"$OUT/canary.stderr" & )
 sleep 8
 win="$(xdotool search --name "Xenia-canary" | tail -1)"
@@ -25,6 +25,6 @@ win="$(xdotool search --name "Xenia-canary" | tail -1)"
 xdotool windowactivate "$win" 2>/dev/null; xdotool key --window "$win" F10; xdotool key F10
 echo "armed at ${SECONDS}s"
 sleep 12; pad "press=A"; sleep 0.4; pad ""; echo "A (skip video) at ${SECONDS}s"
-sleep 45          # let the title build in and settle, no further input
+sleep 150         # let the title build in and settle, no further input
 pkill -x xenia_canary
 echo "done at ${SECONDS}s"; ls -la "$OUT"/*.log 2>/dev/null