Pass 9 post-pass: the uncited surface counted, and it is not where I said

Fix 2 asked for the audit; here it is, and it reorders the fix list.

101 percentages in CLEAN_GROUND.md, 67 of them computed statistics rather
than skill ratings, 14 cited, 53 held by nothing. They split two ways and the
kind finding 1 was about is the safer one: a deterministic figure that drifts
can be caught by anyone who re-derives it in a second.

The exposed surface is EXPOSURE's two pack tables and the 58.7% four-player
figure — 2000-run samples against the declared cast, catchable only by
re-running the exact command, and behind no artifact at all. I had assumed
lethality-baseline.json covered them. It does not: it measures each creature
SOLO against the frozen party holloway/okonkwo/nkemdirim/ferriby, a different
four agents, storing {wipe, down, rounds}. The document's tables read "of 6"
and fight packs of three, six and ten. None of 3.48, 4.79, 5.97, 0.42, 3.03
or 58.7 appears in any baseline in tools/.

Recorded as a decision: derived figures want to resolve against the rule at
check time, since a stored baseline for them is a cache of arithmetic;
sampled figures want a baseline, since re-running them is expensive and
noisy. Same problem, different mechanisms.

Census and the baseline's shape counted here rather than taken from the peer
session that raised the gap.

npm run check: 19 guards pass.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

docs/scenarios/CLEAN_GROUND_PLAYTEST_9.md

@@ -170,3 +170,52 @@ win that, or improvised, or reached for the unsettled read-aloud. There is now a
 
 **Still untested after nine passes:** a fight the players choose; Ashcroft accepted *and*
 taken; a successful swap, now failed four times running; and the human run.
+
+---
+
+## POST-PASS — the uncited surface, counted, and it is not where I said it was
+
+Fix 2 above asked for the other uncited figures to be audited. Done, and **the priority order
+in the fix list is wrong.** Counted in `CLEAN_GROUND.md`:
+
+| | |
+|---|---|
+| Percentages in the document | **101** |
+| Of those, carrying a decimal point — computed statistics rather than skill ratings | **67** |
+| Cited against an artifact | **14** |
+| **Computed statistics with nothing holding them** | **53** |
+
+The 53 are two different kinds of figure, and **the kind finding 1 was about is the safer
+one**:
+
+- **Deterministic** — about sixteen figures: the step-5 splits, the taken/held/unsettled
+  families, the band widths. Enumerable exactly from `opposedContestFor` and `gradeRoll`, no
+  seed, no runs. These drifted, which is finding 1, but they can be re-derived in a second by
+  anybody who suspects them.
+- **Sampled, and held by nothing at all** — **EXPOSURE's two pack tables** and the 58.7%
+  four-player figure. These are 2000-run simulations against the declared cast of six, and
+  **no baseline contains them.**
+
+⚠ **`lethality-baseline.json` does not cover them, which I had assumed it did.** Checked:
+it measures each creature **solo**, against the frozen party
+`pc_holloway, pc_okonkwo, pc_nkemdirim, pc_ferriby` — a different four agents from this
+scenario's cast — and stores `{wipe, down, rounds}` per creature. The understudy's entry is
+`wipe 0, down 0.05, rounds 2`. EXPOSURE's tables read *"of 6"* and fight packs of three, six
+and ten. **Different party, different fights, different shape.** None of 3.48, 4.79, 5.97,
+0.42, 3.03 or 58.7 appears in any baseline in `tools/`.
+
+**So the exposed surface is the opposite of what finding 1 suggested.** A deterministic figure
+that drifts can be caught by anyone who re-derives it. **A sampled figure that drifts can only
+be caught by somebody re-running the exact command against the exact cast**, and the
+scenario's most consequential table — the one that decides whether a GM lets a fight happen —
+is thirty such numbers with no artifact behind them.
+
+**And the two want different mechanisms, not the same one.** A stored baseline for a
+deterministic figure is a cache of the rule, and a guard over it mostly asserts that
+arithmetic has not changed; those should resolve **against the rule at check time**. Sampled
+figures should resolve against a **baseline**, because re-running them is expensive and
+carries noise. Recorded here as a decision rather than left as a preference.
+
+*Census and the `lethality-baseline.json` shape independently verified; the peer session
+raised the pack-table gap and reordered these priorities, and the numbers above were counted
+here rather than taken from that message.*