docs: correct the provenance of the recorded binding defects

An earlier revision claimed every recorded defect was a regression introduced
inside the #737 -> #728 stack, on the grounds that main is green on these
paths. Diffing residual_nodes against main does not support that: the function
is substantively identical there, and the stack changed only how the accuracy
target is derived and the wording of the error message.

main is green because its tests never plan these queries, not because the code
is correct. That is the more worrying reading, so it should be the recorded one.

The fragment mismatch is fixed on main by #781: a context-derived leaf schema
was compared for equality against an isolated re-parse that cannot reproduce
it, and open schemas are now compared by containment.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Author: zzylol

control_plane/tests/issue754_level1.rs

@@ -103,9 +103,16 @@ fn required_summary_shapes(name: &str, fixture: Fixture) -> Vec<(&'static str, R
 /// Binding defects this fixture reaches today. Level 1 records them rather than
 /// tolerating them silently: each entry must still occur exactly `count` times,
 /// so a new instance fails the test and a fixed one forces its line to be
-/// deleted. `main` is green on these paths, so every entry is a regression
-/// introduced inside the #737 → #728 stack, and all of them must be gone before
-/// #775 installs and executes these candidates.
+/// deleted. All of them must be gone before #775 installs and executes these
+/// candidates.
+///
+/// These are *not* regressions introduced by this stack, which is what an
+/// earlier revision of this comment claimed on the grounds that `main` is
+/// green. `residual_nodes` is substantively identical on `main`; the stack
+/// changed only how it derives the accuracy target and the wording of this
+/// message. `main` is green because its tests never plan these queries, not
+/// because the code is correct. The fragment mismatch is fixed on `main` by
+/// #781, which compares context-derived leaf schemas by containment.
 const KNOWN_BINDING_DEFECTS: &[(&str, &str, usize)] = &[
     (
         "grouped-rate",

docs/evaluation/issue754-human-review/README.md

@@ -89,8 +89,11 @@ recorded as a policy reason rather than a defect.
 
 **Binding defects** are bugs. Three candidates fail with `Planner logical
 fragment does not match any original query subtree` and one with a schema
-incompatibility. `main` is green on these paths, so each is a regression
-introduced inside the #737 → #728 stack. They are enumerated in
+incompatibility. They are older than this stack: `residual_nodes` is
+substantively identical on `main`, which is green only because its tests never
+plan these queries. The fragment mismatch is a context-derived leaf schema
+compared for equality against an isolated re-parse that cannot reproduce it;
+#781 fixes it on `main` by comparing those schemas by containment. They are enumerated in
 `KNOWN_BINDING_DEFECTS` in the Level-1 test with exact occurrence counts: a new
 instance fails the test, and a fixed one forces its entry to be deleted. Every
 entry must be gone before #775 installs and executes these candidates. The same