diff --git a/.changeset/eager-closure-header-retired-figures.md b/.changeset/eager-closure-header-retired-figures.md
new file mode 100644
index 0000000000..f45409d3f1
--- /dev/null
+++ b/.changeset/eager-closure-header-retired-figures.md
@@ -0,0 +1,4 @@
+---
+---
+
+Docs-only: the eager-closure gate header stops restating its ceiling and baseline as MiB literals and names the two constants instead, with a pin that refuses a rendered size written back into that note. Internal CI script and its unit test only; nothing published changes.
diff --git a/scripts/__tests__/check-eager-closure-budget.test.ts b/scripts/__tests__/check-eager-closure-budget.test.ts
index 8204d17c43..b754cdd91b 100644
--- a/scripts/__tests__/check-eager-closure-budget.test.ts
+++ b/scripts/__tests__/check-eager-closure-budget.test.ts
@@ -2088,3 +2088,84 @@ describe("chunk counts in this gate's prose (objectui#7528)", () => {
     },
   );
 });
+
+// ── the ceiling note names its constants, and never renders them ─────────────
+
+/**
+ * objectui#8964 — the header note that says WHICH constant is which may name
+ * its two subjects, and may not render either of them as a size.
+ *
+ * The note used to do both. It stated the aggregate ceiling and the baseline as
+ * MiB literals and then explained, correctly and at length, why the two figures
+ * were consistent with each other. A maintainer-authorised re-baseline moved
+ * both constants; the prose stayed. For four days and three subsequent edits to
+ * that file the note rendered the RETIRED pair, explaining the consistency of
+ * two numbers neither of which was in force, and nothing anywhere went red —
+ * because nothing fails on a number written in a comment.
+ *
+ * ⚠️ This is NOT a widening of the chunk-count pin above into a size pin. That
+ * trade was weighed there and declined for a reason that still holds: a general
+ * size pin would have to tell a ceiling's value from the several sizes this
+ * file's prose legitimately carries, and it cannot. This pin does not try. Its
+ * population is two named paragraphs which carry no size at all, and its whole
+ * claim is that they still carry none. The anchored measurements elsewhere in
+ * the header — the re-baseline records, the incident figures — are outside it
+ * and stay exactly as they are.
+ *
+ * ⚠️ Nor may this docblock quote the retired literals back, for the reason the
+ * objectui#7528 block gives: the reader cannot tell a quotation from a claim,
+ * and refusing both is the safe direction. The numbers live in the card.
+ *
+ * The region is located by the sentence that opens it and the sentence that
+ * closes it, both asserted present and unique, so a rewrite that drops either
+ * one turns this red rather than green — an empty scan here would otherwise be
+ * indistinguishable from a note that states nothing.
+ */
+describe('the ceiling note states no rendered size (objectui#8964)', () => {
+  /** The first words of the note, and the last — the region this pin owns. */
+  const OPENS = 'This is a truthful CURRENT-STATE ceiling, not a target.';
+  const CLOSES = 'rendered size written back into either paragraph.';
+
+  /** A size as a person writes one: a numeral, then a byte unit. */
+  const RENDERED_SIZE = /\b\d[\d,_]*(?:\.\d+)?\s*(?:[KMGT]i?B)\b/g;
+
+  const renderedSizes = (text: string): string[] =>
+    [...text.matchAll(RENDERED_SIZE)].map((m) => m[0].replace(/\s+/g, ' ').trim());
+
+  /** The note, as the pin reads it. Throws rather than returning nothing. */
+  function ceilingNote(source: string): string {
+    const start = source.indexOf(OPENS);
+    const end = source.indexOf(CLOSES);
+    if (start < 0 || end < 0) throw new Error('the ceiling note is not where this pin looks for it');
+    return source.slice(start, end + CLOSES.length);
+  }
+
+  it('reads a note that is there exactly once, rather than reading nothing', () => {
+    const source = fs.readFileSync(checkerPath, 'utf8');
+    expect(source.split(OPENS).length - 1).toBe(1);
+    expect(source.split(CLOSES).length - 1).toBe(1);
+    // A floor, so a note shrunk to its two anchors cannot pass by carrying nothing.
+    expect(ceilingNote(source).length).toBeGreaterThan(800);
+  });
+
+  it('sees a rendered size when one is in front of it', () => {
+    expect(renderedSizes('a 9.99 MB payload, 512 KiB of it new, over a 7 GB disk')).toEqual([
+      '9.99 MB',
+      '512 KiB',
+      '7 GB',
+    ]);
+    expect(renderedSizes('objectui#8964 names two constants and renders neither')).toEqual([]);
+  });
+
+  it('renders neither constant as a size', () => {
+    const rendered = renderedSizes(ceilingNote(fs.readFileSync(checkerPath, 'utf8')));
+    expect(
+      rendered,
+      `the note that says which constant is which renders a size (${rendered.join(', ')}). ` +
+        'A figure written here is a second copy of a constant that lives a few lines below it, and ' +
+        'only one of the two moves when the ceiling is re-baselined — which is how this note came to ' +
+        'explain, for four days, why two retired numbers were consistent with each other. Name the ' +
+        'constant and let the gate print the reading (objectui#8964).',
+    ).toEqual([]);
+  });
+});
diff --git a/scripts/check-eager-closure-budget.mjs b/scripts/check-eager-closure-budget.mjs
index 9b5e9a7798..84b9ea7849 100644
--- a/scripts/check-eager-closure-budget.mjs
+++ b/scripts/check-eager-closure-budget.mjs
@@ -205,20 +205,31 @@
  * become an excuse to widen it — a ceiling that rises while the sensitivity
  * relaxes is a gate quietly retiring itself.
  *
- * This is a truthful CURRENT-STATE ceiling, not a target. 3.07 MB gzipped
- * before first render is a bad payload, and the honest long-term line is far
- * below it — but lowering the line to a TARGET is a separate decision with its
- * own work behind it (objectui#5324 names the candidates), and re-baselining
- * onto a fresh measurement is not that. Nothing here should be read as a
- * finding that 3.12 MB is acceptable.
- *
- * The two figures in that paragraph are one constant each, rendered in MiB, and
- * not a contradiction: 3.07 MB is {@link BASELINE}'s measured payload and
- * 3.12 MB is {@link MAX_EAGER_CLOSURE_GZIP_BYTES}, the ceiling standing over it.
- * Saying which is which is the whole of this note — a paragraph that names two
- * sizes without naming their subjects reads as one number that changed its mind
+ * This is a truthful CURRENT-STATE ceiling, not a target. A multi-megabyte
+ * gzipped payload before first render is a bad payload, and the honest
+ * long-term line is far below it — but lowering the line to a TARGET is a
+ * separate decision with its own work behind it (objectui#5324 names the
+ * candidates), and re-baselining onto a fresh measurement is not that. Nothing
+ * here should be read as a finding that the ceiling standing over the closure
+ * today is acceptable.
+ *
+ * Which is which: {@link BASELINE} carries the measured payload and
+ * {@link MAX_EAGER_CLOSURE_GZIP_BYTES} is the ceiling standing over it. Saying
+ * which is which is the whole of this note — a paragraph that names two sizes
+ * without naming their subjects reads as one number that changed its mind
  * (objectui#7528).
  *
+ * ⛔ Neither paragraph renders either constant as a size, deliberately. This
+ * note used to state both of them as MiB literals; a re-baseline moved both
+ * constants, and four days and three later edits to this file went by with the
+ * RETIRED pair still written here and nothing red anywhere across that
+ * distance, because nothing fails on a number written in a comment
+ * (objectui#8964). The subjects are named and the readings are left to the
+ * instrument: the two constants below carry them in bytes, each pinned to the
+ * commit it was measured on, and the gate's verdict line prints the live one on
+ * every run. `scripts/__tests__/check-eager-closure-budget.test.ts` refuses a
+ * rendered size written back into either paragraph.
+ *
  * ## Per-chunk ceilings (objectui#5490)
  *
  * One total over the whole eager closure cannot say WHERE the payload moved, and
@@ -255,8 +266,9 @@
  *
  * Re-baselining is legitimate — it is how a ratchet advances — but it is a
  * DECISION, so make it visible: update the constant, update the measured figure
- * in this comment, and say in the PR what the added bytes buy. Silently bumping
- * the number to make CI green reproduces the gate this file replaced.
+ * in the prose attached to it, and say in the PR what the added bytes buy.
+ * Silently bumping the number to make CI green reproduces the gate this file
+ * replaced.
  */
 
 import fs from 'node:fs';