diff --git a/.github/workflows/lint.yml b/.github/workflows/lint.yml
index bb0c37ed2f..022f50a8e5 100644
--- a/.github/workflows/lint.yml
+++ b/.github/workflows/lint.yml
@@ -1081,9 +1081,12 @@ jobs:
       # — one card per silent exit — and each fix lands as more unexercised
       # code. This step is the exercise: the shipped script is extracted from
       # the YAML with a real parser (never retyped) and run under doubles the
-      # way actions/github-script runs it, as one AsyncFunction body. Ten
+      # way actions/github-script runs it, as one AsyncFunction body. Its
       # scenarios pin the target parse and the outcome of EVERY exit — which of
       # setFailed / warning / job summary fires, and which API calls were made.
+      # How many of them there are is deliberately NOT written here (#9917):
+      # the check names the total on its own OK line, and `--list` prints the
+      # whole table. This comment used to say "ten" against a real eighteen.
       #
       # Assertion 0 is the compile, and it is not theoretical: this job failed
       # twice on 2026-08-02 with `SyntaxError: Identifier 'octokit' has already
@@ -1091,14 +1094,26 @@ jobs:
       # workflow whose red nothing else in CI can see.
       #
       # Its --self-test runs first and is the half that stops the battery
-      # rotting into decoration: it mutates the shipped script seven ways —
-      # downgrade the verdict to a warning, stop collecting failed keys, break
-      # out of the loop instead of isolating, drop the same-repo skip, narrow
-      # the keyword set, drop the already-closed skip, downgrade the notice
-      # path's verdict — and requires the battery to go RED for each, naming the
-      # scenario that catches it. A mutation whose anchor no longer exists is a
-      # failure too, so a rewrite of the workflow cannot leave the mutations
-      # silently matching nothing.
+      # rotting into decoration: it mutates the shipped script along every
+      # CLASS the contract is made of — a verdict downgraded (red becomes a
+      # warning, or an info line nothing annotates), the loop's bookkeeping
+      # deleted (which keys failed, which half was lost, which target was
+      # refused), isolation abandoned mid-loop, the target parse narrowed (the
+      # keyword set, the optional colon, the same-repo skip), a guard removed
+      # (already-closed, triage, pull-request), idempotency eroded (the
+      # backlink marker dropped, or a blind post where a skip belongs) — and
+      # requires the battery to go RED for each, naming the scenario that
+      # catches it. A mutation whose anchor no longer exists is a failure too,
+      # so a rewrite of the workflow cannot leave the mutations silently
+      # matching nothing.
+      #
+      # Classes rather than a count, and rather than the enumeration that used
+      # to sit here, because BOTH hand-maintained numbers in this comment had
+      # gone stale unnoticed — measured 2026-08-20 from the check's own output,
+      # eighteen scenarios and fifteen mutations against a comment claiming ten
+      # and seven (#9917). A stale enumeration reads exactly as authoritative
+      # as a fresh one, and this comment is what the next reader checks the
+      # harness against. Prose classes survive an M16; a count does not.
       #
       # Invoked as `node` rather than through a `pnpm check:*` alias: that alias
       # belongs in root package.json, declared territory of the @changesets/cli
diff --git a/scripts/check-cross-repo-closer-outcome.mjs b/scripts/check-cross-repo-closer-outcome.mjs
index 2429d31740..826d61b3f1 100644
--- a/scripts/check-cross-repo-closer-outcome.mjs
+++ b/scripts/check-cross-repo-closer-outcome.mjs
@@ -89,15 +89,26 @@
 //
 // A battery of assertions over a script that is already correct is green on day
 // one and green forever, including the day someone deletes the thing it
-// guards. `--self-test` mutates the extracted source -- drop the `setFailed`,
-// stop collecting failed keys, `break` out of the loop instead of isolating,
-// disable the same-repo skip, narrow the keyword set, collapse the
-// already-closed branch, strip the backlink marker, drop the triage guard,
-// post blind when the comment listing is refused, stop recording which half
-// was lost -- and requires the battery to go RED for each, naming the scenario
-// it expects. Each mutation also asserts its own anchor was PRESENT before
-// substituting: a mutation that silently matched nothing would leave the
-// battery green and read exactly like a passing self-test.
+// guards. `--self-test` mutates the extracted source and requires the battery
+// to go RED for each mutation, naming the scenario it expects.
+//
+// What the mutations cover is written here as CLASSES, and their number is not
+// written here at all (#9917). The list this paragraph used to carry named ten
+// of them and was never updated again; the same habit in lint.yml's step
+// comment was counting seven when there were fifteen. A prose class survives
+// the next mutation being added, a hand-count does not, and `--self-test`
+// prints its own total on every run. The classes: a verdict DOWNGRADED (red
+// becomes a warning, or an info line nothing annotates), the loop's
+// BOOKKEEPING deleted (which keys failed, which half was lost, which target
+// was refused -- each one lets a run report green about work it did not do),
+// ISOLATION abandoned mid-loop, the TARGET PARSE narrowed (the keyword set,
+// the optional colon, the same-repo skip), a GUARD removed (already-closed,
+// triage, pull-request), and IDEMPOTENCY eroded (the backlink marker dropped,
+// or a blind post where a skip belongs).
+//
+// Each mutation also asserts its own anchor was PRESENT before substituting: a
+// mutation that silently matched nothing would leave the battery green and
+// read exactly like a passing self-test.
 //
 // One scenario (L11) is driven TWICE, the second run's world built out of the
 // first run's calls. Re-run idempotency is a property of the PAIR, and a
@@ -200,9 +211,14 @@ function makeDoubles({ body, token, issues = {}, prCommentError = null, summaryE
       calls.get.push(`${owner}/${repo}#${n}`);
       const t = target(owner, repo, n);
       if (t.getError) throw t.getError;
-      // `state_reason` is nullable on a real closed issue -- objectui#4478
-      // answers `null` from this very endpoint -- so the default models that
-      // rather than inventing a value the API does not promise.
+      // `state_reason` is nullable BY CONTRACT on this endpoint -- the API
+      // states a reason when it has one and promises nothing otherwise -- so
+      // the default models `null` rather than inventing a value the API does
+      // not promise. Modelled, not sampled, and that distinction is load
+      // bearing here: the citation that stood on this line named a foreign
+      // number as a closed issue answering `null` and it was a pull request
+      // (#9917). L9 carries the reason a sampled specimen is the wrong kind of
+      // evidence for THIS default in particular.
       //
       // `pull_request` is how the same endpoint says the number is a PULL
       // REQUEST, and it is modelled as ABSENCE rather than as `undefined`
@@ -667,9 +683,25 @@ export const SCENARIOS = [
     scenario: () => ({
       body: MIXED_BODY,
       token: 'pat',
-      // Measured, not invented: objectstack-ai/objectui#4478 is closed and
-      // answers `state_reason: null` from `issues.get`. A fix that keys on the
-      // reason has to say what null means, and this pins the answer.
+      // MODELLED FROM THE CONTRACT, deliberately not sampled. `issues.get`
+      // declares `state_reason` nullable and promises no reason on a closed
+      // issue, so a fix that keys on the reason has to say what null means, and
+      // this pins the answer -- the shipped script's own reading of it, "no
+      // objection recorded", is stated on the already-closed branch of
+      // .github/workflows/cross-repo-issue-closer.yml, which is the subject
+      // this scenario drives.
+      //
+      // A SAMPLED citation stood here and was withdrawn (#9917): it named
+      // objectui#4478 as a closed issue answering `null`, and that number is a
+      // PULL REQUEST. Read that as the trap it is rather than as one bad
+      // lookup -- the objects that most dependably answer `state: 'closed'`
+      // with `state_reason: null` ARE pull requests, so hunting a specimen for
+      // this fixture walks straight into L13's subject, and the file then
+      // offers ONE observation as evidence for two opposite scenarios. L9 and
+      // L13 are kept on separate footings on purpose: L9 models the contract
+      // and cites nothing, L13 cites a measured pull request
+      // (objectstack-ai/objectstack#9143). Do not "improve" this by finding a
+      // real issue to name.
       issues: { [CLOSED_TARGET]: { state: 'closed', stateReason: null, comments: [] } },
     }),
     check: (r, t) => [
@@ -845,10 +877,13 @@ export const SCENARIOS = [
       token: 'pat',
       // A merged pull request answers `state: 'closed'` with
       // `state_reason: null` from the issues endpoint -- measured on
-      // objectstack-ai/objectstack#9143. That is the very fixture L9 uses to
-      // prove a closed ISSUE gets its backlink, which makes this scenario the
-      // ORDERING test: a `pull_request` guard placed after the state branch
-      // would sail past it and comment on somebody else's pull request.
+      // objectstack-ai/objectstack#9143. That is the very fixture SHAPE L9 uses
+      // to prove a closed ISSUE gets its backlink -- the shape, not the
+      // observation: L9 models its null from the contract and cites no object,
+      // precisely so that this measured pull request stays evidence for THIS
+      // scenario only (#9917). The shared shape is what makes this the ORDERING
+      // test: a `pull_request` guard placed after the state branch would sail
+      // past it and comment on somebody else's pull request.
       issues: {
         [PR_TARGET]: {
           state: 'closed',