diff --git a/docs/noise/vulcanus-cliffs-NOTES.md b/docs/noise/vulcanus-cliffs-NOTES.md
index 9f05166..62d2f1e 100644
--- a/docs/noise/vulcanus-cliffs-NOTES.md
+++ b/docs/noise/vulcanus-cliffs-NOTES.md
@@ -996,3 +996,58 @@ sides of a comparison ran different **rule sets** - and in both cases the
 mismatch had a plausible mechanism ready to absorb it, so the wrong explanation
 was the comfortable one. Before attributing a residual to a mechanism, check
 that both sides of the comparison are running the same rules.
+
+## Channel audit: cliffs really are the only coarse-grid consumer (#84 item 4)
+
+#83 left an assumption behind: cliffs are the only consumer reading a
+`multisample`-bearing field on a grid other than 1 tile. That was never
+measured - which is precisely the status the cliff case itself had before #83 -
+so it is measured now. `test/multisampleChannelAudit.spec.ts`.
+
+**The surface is much smaller than "audit every consumer", and that is the first
+finding.** `multisample` appears in `~/GitHub/factorio-data` @ 2.1.12 in exactly
+**one** place, `vulcanus_basalt_lakes_multisample`
+(`planet-vulcanus-map-gen.lua:547`), used only by `vulcanus_elev`. No other
+planet uses the primitive at all. So the audit reduces to "who consumes Vulcanus
+`elevation`":
+
+| consumer | reads `elevation`? | grid | how established |
+| --- | --- | --- | --- |
+| cliff generator | yes, via `cliff_elevation_from_elevation` | **4** | measured, #83 |
+| tile generator (19 `*_range` expressions) | yes, via `vulcanus_elev` | **1** | measured, below |
+| `vulcanus_temperature` | yes, `min(elev, elev/100)` | 1 | same program as tiles |
+| resources: tungsten, coal, calcite, sulfuric acid | **no** | n/a | read below |
+| rocks, geyser | **no** | n/a | read below |
+
+**The resource row is the one that had to be checked rather than assumed**,
+because the generic path *does* couple to elevation:
+`starting_resources_lake_mask = clamp((elevation - 1) / 10, 0, 1)` feeds
+`starting_patches`' `spot_favorability_expression`
+(`core/prototypes/noise-programs.lua:270`), and CLAUDE.md flags that this
+coupling is exactly what changed at 2.1.9. Vulcanus does not take that path -
+each of its four resources is placed by its own `vulcanus_*_region` expression,
+and those four definitions contain **0** references to `elevation`. So the lake
+mask is a Nauvis-only coupling and no Vulcanus resource can see the multisample.
+
+### The tile generator reads grid 1, measured against the game
+
+Substituting the cliff generator's 4-tile elevation into the tile resolver and
+comparing against the game's own `get_tile` output over the 381 oracle
+positions:
+
+| | grid 1 (ships) | grid 4 |
+| --- | --- | --- |
+| tile-name agreement | **0.9816** | 0.8609 |
+| lava misclassifications | **0 / 381** | 27 / 381 |
+
+46 tiles would be named wrongly, and the binary lava call - the only thing the
+cliff collision rejection reads - goes from exact to 27 wrong. The 1-tile field
+is not merely adequate here; it is the one that matches, and the metric is
+demonstrably sensitive to the swap. That sensitivity is the part worth insisting
+on: PR #57's substitution failed precisely because "nothing changed" was
+indistinguishable from "the substitution never ran".
+
+**A side result.** This also clears the lava perimeter that costs 13 real cliffs
+their placement: reading the other elevation channel makes the lava
+classification dramatically worse, not better, so the perimeter error is not a
+channel mistake. It is somewhere else in `vulcanusCatalog`.
diff --git a/test/multisampleChannelAudit.spec.ts b/test/multisampleChannelAudit.spec.ts
new file mode 100644
index 0000000..1b42882
--- /dev/null
+++ b/test/multisampleChannelAudit.spec.ts
@@ -0,0 +1,152 @@
+import { describe, expect, it } from "vite-plus/test";
+
+import fixture from "./fixtures/oracle-vulcanus-tile-names.seed123456.json";
+import { VULCANUS_CLIFF_BLOCKING_TILES } from "../src/noise/preview/renderVulcanusCliffs";
+import {
+  makeVulcanusStack,
+  makeVulcanusTileResolverFrom,
+  type VulcanusStack,
+} from "../src/noise/tiles/vulcanusCatalog";
+
+/**
+ * **The channel audit for `multisample` (issue #84 item 4).**
+ *
+ * #83 found that `multisample`'s offsets are in the CALLING noise program's grid
+ * units, so `vulcanus_basalt_lakes_multisample` is a 4-tile min-filter for the
+ * cliff generator and a 1-tile one for `calculate_tile_properties`. The port had
+ * used the 1-tile field for both. The obvious follow-up question is whether any
+ * OTHER consumer reads a `multisample`-bearing field on a grid that is not 1
+ * tile - and cliffs being the only coarse-grid consumer was an assumption, not a
+ * measurement, which is exactly the status the cliff case had before #83.
+ *
+ * **The audit surface is far smaller than it looks, and that is a finding.**
+ * `multisample` appears in `~/GitHub/factorio-data` @ 2.1.12 in exactly ONE
+ * place - `vulcanus_basalt_lakes_multisample`
+ * (`space-age/prototypes/planet/planet-vulcanus-map-gen.lua:547`), used only by
+ * `vulcanus_elev`. Nothing on Nauvis, Gleba, Fulgora or Aquilo uses the
+ * primitive at all. So "audit every consumer of a multisample-bearing field"
+ * reduces to "audit every consumer of Vulcanus `elevation`", and that list is
+ * short enough to close by reading the planet definition
+ * (`space-age/prototypes/planet/planet-map-gen.lua:4-30`):
+ *
+ * | consumer | reads `elevation`? | grid |
+ * | --- | --- | --- |
+ * | cliff generator (`cliff_elevation_from_elevation`) | yes | **4** - #83 |
+ * | tile generator (the 19 `*_range` expressions) | yes, via `vulcanus_elev` | 1 - measured below |
+ * | `vulcanus_temperature` | yes, `min(elev, elev/100)` | 1, same program as tiles |
+ * | resources (tungsten / coal / calcite / sulfuric acid) | **no** | n/a |
+ * | rocks, geyser | **no** | n/a |
+ *
+ * The resource line is the one worth stating explicitly, because the generic
+ * `resource_autoplace_all_patches` path DOES couple to elevation -
+ * `starting_resources_lake_mask = clamp((elevation - 1) / 10, 0, 1)` feeds
+ * `starting_patches`' `spot_favorability_expression`
+ * (`core/prototypes/noise-programs.lua:270`). Vulcanus does not use that path:
+ * each of its four resources is placed by its own `vulcanus_*_region`
+ * expression, and grepping those four definitions for `elevation` returns **0**.
+ * So the lake mask is a Nauvis-only coupling and no Vulcanus resource can see
+ * the multisample. (`src/noise/resources/startingPatches.ts` hardcodes
+ * `makeElevationNauvis` for the same reason.)
+ *
+ * That leaves the tile generator as the only consumer whose grid had to be
+ * measured rather than reasoned about, which is what this file does.
+ */
+describe("multisample channel audit - which consumers read which grid", () => {
+  const positions = fixture.positions;
+  const want = fixture.tileNames;
+
+  /** A stack whose tile-facing `elev` is the CLIFF generator's 4-tile field. */
+  const swappedStack = (base: VulcanusStack): VulcanusStack => ({
+    ...base,
+    elevation: {
+      ...base.elevation,
+      elev: (x, y) => base.elevation.cliffElevation(x, y),
+      elevation: (x, y) => base.elevation.cliffElevation(x, y),
+    },
+  });
+
+  const agreement = (resolve: (x: number, y: number) => { name: string }): number => {
+    let agree = 0;
+    for (let i = 0; i < positions.length; i++)
+      if (resolve(positions[i].x, positions[i].y).name === want[i]) agree++;
+    return agree / positions.length;
+  };
+
+  /**
+   * The substitution is only exact if the `max(-500, ...)` clamp never bites,
+   * because `cliffElevation` is clamped and `elev` is not. Asserted rather than
+   * assumed: if a probe position ever sits below -500 this arm is comparing
+   * something subtly different from "the same field at grid 4".
+   */
+  it("the clamp never bites at these positions, so the swap is exact", () => {
+    const stack = makeVulcanusStack({ seed0: fixture.seed0 });
+    let min = Infinity;
+    for (const p of positions) min = Math.min(min, stack.elevation.elev(p.x, p.y));
+    expect(min).toBeGreaterThan(-500);
+  }, 120000);
+
+  /**
+   * **The tile generator runs at grid 1, measured.** Feeding it the cliff
+   * generator's 4-tile elevation instead collapses agreement with the game's own
+   * `get_tile` output. The 1-tile field is not merely adequate - it is the one
+   * that matches, and the metric is demonstrably sensitive to the difference,
+   * which is what stops "no change" from being the vacuous answer it was for
+   * PR #57's field substitution.
+   *
+   * Measured 2026-08-01 over the 381 oracle positions: **0.9816 at grid 1
+   * against 0.8609 at grid 4** - 46 tiles the port would name wrongly if the
+   * tile generator shared the cliff generator's channel. The gap is the whole
+   * point; the grid-4 number is pinned only loosely, as an upper bound well
+   * below the grid-1 floor.
+   */
+  it("TILES read the 1-tile channel - the 4-tile field disagrees with the game", () => {
+    const base = makeVulcanusStack({ seed0: fixture.seed0 });
+    const atGrid1 = agreement(makeVulcanusTileResolverFrom(base));
+    const atGrid4 = agreement(makeVulcanusTileResolverFrom(swappedStack(base)));
+
+    // Grid 1 is the shipping configuration and `vulcanusTiles.spec.ts` owns its
+    // floor; repeated here only so the comparison is self-contained.
+    console.log(`vulcanus tile agreement: grid1=${atGrid1.toFixed(4)} grid4=${atGrid4.toFixed(4)}`);
+    expect(atGrid1).toBeGreaterThan(0.978);
+    // The substitution must actually move the answer. Without this, a swap that
+    // silently failed to apply would read as "the grids agree".
+    expect(atGrid4).toBeLessThan(atGrid1 - 0.02);
+  }, 120000);
+
+  /**
+   * The same question asked of the binary lava classification, which is the only
+   * thing the cliff collision rejection reads. It is exact at grid 1 (49/49 both
+   * ways, pinned in `vulcanusTiles.spec.ts`); at grid 4 it is **27 wrong out of
+   * 381**.
+   *
+   * This is the sharper of the two arms for the audit's purpose. A tile-name
+   * argmax can flip between two members of one biome family for precision
+   * reasons, but lava/not-lava is a wide margin, so a mismatch here means the
+   * field is genuinely different rather than near a tie. 0 -> 27 is that.
+   *
+   * It also settles a question the cliff work left hanging in the other
+   * direction: the lava perimeter that costs 13 real cliffs their placement
+   * (`vulcanusCliffEntities.spec.ts`) is NOT the tile resolver reading the wrong
+   * elevation channel. Reading the other channel makes lava dramatically worse,
+   * not better.
+   */
+  it("LAVA classification is exact at grid 1 and not at grid 4", () => {
+    const base = makeVulcanusStack({ seed0: fixture.seed0 });
+    const count = (resolve: (x: number, y: number) => { name: string }): number => {
+      let mismatch = 0;
+      for (let i = 0; i < positions.length; i++) {
+        const ours = VULCANUS_CLIFF_BLOCKING_TILES.has(
+          resolve(positions[i].x, positions[i].y).name,
+        );
+        if (ours !== VULCANUS_CLIFF_BLOCKING_TILES.has(want[i])) mismatch++;
+      }
+      return mismatch;
+    };
+    const at1 = count(makeVulcanusTileResolverFrom(base));
+    const at4 = count(makeVulcanusTileResolverFrom(swappedStack(base)));
+    console.log(`vulcanus lava mismatches: grid1=${String(at1)} grid4=${String(at4)} of 381`);
+    expect(at1).toBe(0);
+    // Non-vacuity in the same breath: the swap must actually change the field.
+    expect(at4).toBeGreaterThan(10);
+  }, 120000);
+});