test(retrieval-gate): red -- one real set of three is not a measurement of three

PM's checkpoint on 2c8296b ran eight cheating attacks at this gate and four
went through: a row came back GREEN without one label becoming true or one
concept ranking better. This is the first of them, written as a test that must
refuse it.

J2: `row_eight` iterates over whatever `--real` handed it and counts. One set
of three came back `6 of 6 GREEN`, and the realistic route is `--real r761`
alone -- the set PM's own registration puts at 7 of 7, with the two that miss
left out. The docstring already said "never green by leaving a set out"; the
code did not, and THIS REPOSITORY'S OWN TEST asserted the opposite at
`tests/test_retrieval_gate.py:486`, `(1, 1, GREEN)` for a single wiki set.
That assertion is corrected here rather than worked around: a test that pins
the defect is the defect.

J2b: the headline `k of N` is `quoted_hits + concept_hits` over
`quoted_units + concept_units`, written on the line directly above the detail
that says the two granularities are NOT summed into one number.

3 red on an assertion about behaviour (GREEN where NOT RUN is required, and a
headline that is a sum), 46 passed. The green direction -- all three named
sets present -- is in the same commit and passes already, so the fix cannot
close the rows by making the row unreachable.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

tests/test_retrieval_gate.py

@@ -483,7 +483,9 @@ def test_the_wiki_adapter_reads_its_own_shape_and_hits_by_source_file(tmp_path:
     assert question_set.questions[0].fasit[0].by == "source_file"
     assert question_set.questions[0].fasit[0].value == "haandbok.md"
     row = gate.row_eight([(question_set, {"wiki": _bundles(tmp_path)["positive"]})])
-    assert (row.k, row.m, row.status) == (1, 1, gate.GREEN)
+    # J2: one set of three is NOT a measurement of the three -- this asserted
+    # GREEN until 2026-09-19, which is the breakthrough PM measured.
+    assert row.status == gate.NOT_RUN
     assert any("citation granularity" in detail for detail in row.details)
 
 
@@ -736,3 +738,79 @@ def _mapping_is_sorted(mapping: Mapping[str, object]) -> bool:
 def test_every_class_is_documented_in_the_output() -> None:
     assert [letter for letter, _ in gate.CLASSES] == ["a", "b", "c", "d", "e"]
     assert all(description for _, description in gate.CLASSES)
+
+
+# --- the eight attacks PM ran against this gate 2026-09-19 ---------------------
+#
+# Four of them went through: a row went GREEN without one label becoming true or
+# one concept ranking better. Each is reproduced here as a test that must REFUSE
+# it, and the four that were already refused stay as regression guards, so the
+# table is 8 of 8 rather than 4 of 4.
+
+
+def _hitting_set(set_id: str, *, quote: str) -> gate.QuestionSet:
+    """A set the positive bundle answers, named by concept: the shortest way to
+    drive row 8 green without an adapter between the set and the row."""
+    return gate.QuestionSet(
+        set_id=set_id,
+        bundle="positive",
+        path=Path(f"/nowhere/{set_id}.json"),
+        sha256="0" * 64,
+        questions=(
+            gate.Question(
+                id=f"{set_id}-1",
+                question="Naar kontrolleres vinterberedskapen paa hytta?",
+                fasit=(
+                    gate.Fasit(
+                        by="concept", value="haandbok/vinterberedskap", quote=quote
+                    ),
+                ),
+            ),
+        ),
+        controls=(),
+    )
+
+
+def test_j2_row_eight_is_not_run_when_a_required_set_is_left_out(tmp_path: Path) -> None:
+    """PM's J2: one set of three came back `6 of 6 GREEN`."""
+    bundles = {"positive": _bundles(tmp_path)["positive"]}
+    row = gate.row_eight([(_hitting_set("wiki-20", quote="innen 1. november"), bundles)])
+    assert row.status == gate.NOT_RUN
+    assert row.fails
+    assert "r761-sk2" in row.reason and "vegnormal-32" in row.reason
+    # The numbers it DID measure are still carried: a missing set must not cost
+    # the reader the set that ran.
+    assert any("wiki-20: " in detail for detail in row.details)
+
+
+def test_row_eight_is_green_only_with_all_three_named_sets(tmp_path: Path) -> None:
+    bundles = {"positive": _bundles(tmp_path)["positive"]}
+    real = [
+        (_hitting_set(name, quote="innen 1. november"), bundles)
+        for name in ("wiki-20", "r761-sk2", "vegnormal-32")
+    ]
+    row = gate.row_eight(real)
+    assert row.status == gate.GREEN
+    assert (row.k, row.m) == (3, 3)
+
+
+def test_j2b_row_eight_never_sums_the_two_granularities_into_its_headline(
+    tmp_path: Path,
+) -> None:
+    """PM's J2b: the headline was `quoted + concept` on the line above a detail
+    saying the two are not summed."""
+    bundles = {"positive": _bundles(tmp_path)["positive"]}
+    real = [
+        (_hitting_set("wiki-20", quote="innen 1. november"), bundles),
+        (_hitting_set("r761-sk2", quote=""), bundles),
+        (_hitting_set("vegnormal-32", quote=""), bundles),
+    ]
+    row = gate.row_eight(real)
+    # Three questions, one per set: the headline is at QUESTION granularity and
+    # is never the sum of one citation unit and two concept units.
+    assert (row.k, row.m) == (3, 3)
+    assert "question" in row.reason
+    assert any(
+        "1 of 1 at citation granularity, 2 of 2 at concept granularity" in detail
+        for detail in row.details
+    )