volt: path_claim_check — docstring carries no census figures (the first draft's moved with the matcher fix)

tools/path_claim_check.py

@@ -12,14 +12,20 @@ WHY THIS EXISTS
   rule is that a reader must be able to go and look. A path a reader cannot
   open is a citation that fails the only test a stranger can run.
 
-  The measurement that made this tool possible also bounded it: over the whole
-  corpus at HEAD, 1,024 distinct path tokens appear, and only 317 of them are
-  even candidates for "this archive holds that file". The other 707 begin with
-  a component this repo does not track -- scratch trees (`after/verify.txt`),
-  host artifacts (`/root/crucible/...`), other repositories, served routes and
-  URLs. Grading those as missing would be 707 false defects, so they are COUNTED
-  and never graded. What is left is small enough to read: 8 missing paths, each
-  of which is in the reviewed ledger with the class that explains it.
+  The measurement that made this tool possible also bounded it. Over the whole
+  corpus, most distinct path tokens are not candidates for "this archive holds
+  that file" at all: they begin with a component this repo does not track --
+  scratch trees (`after/verify.txt`), host artifacts (`/root/crucible/...`),
+  other repositories, served routes and URLs. Grading those as missing would be
+  hundreds of false defects, so they are COUNTED and never graded. What is left
+  is small enough for a reader: every MISSING path at HEAD is in
+  `tools/path_claim_baseline.tsv` with the class that explains it.
+
+  This docstring deliberately carries no census figures. The first draft did,
+  and the leading-slash fix in the matcher moved every one of them (1,024
+  distinct tokens became 1,036) -- the defect class this archive grades as a
+  stale claim. The tool prints its own census on every run; that print is the
+  only place those numbers live.
 
 WHAT IT MEASURES -- one verdict per DISTINCT path (unit: distinct paths)
   EXISTS    the path is in the tree at --rev.