fix(tooling): derive check-wildcard-fallthrough's self-test case count from its own ledger

The self-test verdict printed a transcribed `17 cases` while the body ran 18
assertions and the battery roster pinned 18 — a number nothing derived and
nothing compared. The literal was left wrong deliberately when the floor
landed, so that change could prove byte-identical output; correcting it is
this change.

The line now sums `batterySeen`, the same ledger the floor immediately above
evaluates, so the printed number and the floor can no longer disagree and a
block that stops running shrinks the count instead of leaving a stale literal
standing. What it counts is assertions that RAN, not `assert(` call sites in
the source — today both are 18, and the comment says which one is meant.

The in-place comment that explained why the literal was knowingly wrong, and
the roster's present-tense description of the verdict as transcribed, are both
rewritten: left standing next to a derived line they would tell the next
reader the opposite of what the file does.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vbw3RPgdtqesx4azk9SbW8

Author: claude

scripts/check-wildcard-fallthrough.mjs

@@ -460,9 +460,11 @@ const SELF_TEST_VERDICT = 'check-wildcard-fallthrough self-test reached its verd
 //
 // "no assertion threw" used to be this self-test's ONLY success condition, so
 // "every case held" and "the cases never ran" printed the same line — and this
-// file's verdict spells a TRANSCRIBED count (`17 cases`), which is evidence, not
-// proof: nothing compared it, so a deleted block shrank the real count while the
-// literal stayed put. Closed the way PR #13487 validated on check-doc-authoring:
+// file's verdict once spelled a TRANSCRIBED count (a literal `17` against a real
+// 18), which is evidence, not proof: nothing compared it, so a deleted block
+// shrank the real count while the literal stayed put. The verdict now derives
+// its number from the ledger this roster floors (#15231). Closed the way PR
+// #13487 validated on check-doc-authoring:
 // what is pinned is the registered NAMES, not a number, and the floor requires
 // the OPENED set to equal the DECLARED set with each battery at or above its own
 // count.
@@ -625,13 +627,16 @@ function selfTest() {
     process.exit(1);
   }
 
-  // ⚠️ MEASURED at 18, not 17: this transcribed literal had already drifted one
-  // below the real count before this floor existed, which is precisely why a
-  // printed number is evidence and not proof. It is left as-is here so this
-  // change stays a pure no-op on output; the correction is filed separately, and
-  // the floor above is what makes the drift harmless — the count can no longer
-  // shrink in silence.
-  console.log('✓ self-test: 17 cases');
+  // The verdict DERIVES its number from `batterySeen` — the very ledger the floor
+  // above just evaluated — so the printed count and the floor can never disagree,
+  // and a block that stops running shrinks the number instead of leaving a
+  // transcribed literal standing (#15231). What it counts is assertions that
+  // actually RAN this run, not `assert(` call sites in the source: those are two
+  // different facts, and only the first one is measured here. Summing the whole
+  // ledger is exact because reaching this line means every battery that
+  // registered is a declared one — the set difference above reds otherwise.
+  const casesRun = [...batterySeen.values()].reduce((total, count) => total + count, 0);
+  console.log(`✓ self-test: ${casesRun} cases`);
 
   return SELF_TEST_VERDICT;
 }