diff --git a/core/capability/reporting.py b/core/capability/reporting.py
index 9bd30d6e..7582feb8 100644
--- a/core/capability/reporting.py
+++ b/core/capability/reporting.py
@@ -43,7 +43,7 @@ _FLAG_CATALOG: dict[str, dict[str, str]] = {
     "transitive_surface": {"state": "flag_shipped_default_off", "adr": "ADR-0083"},
     "gloss_aware_cause": {"state": "flag_shipped_default_off", "adr": "ADR-0085"},
     "thread_anaphora": {"state": "flag_shipped_default_off", "adr": "P3.2"},
-    "discourse_planner": {"state": "flag_shipped_default_off", "adr": "ADR-0089"},
+    "discourse_planner": {"state": "flag_shipped_default_on", "adr": "ADR-0089"},
     "compound_intent_dispatch": {"state": "substrate_shipped_flag_missing", "adr": "ADR-0089-C2"},
     "inference_trace": {"state": "substrate_missing", "adr": "ADR-0024"},
 }
@@ -355,6 +355,7 @@ def chain_report() -> dict[str, Any]:
 def flag_report() -> dict[str, Any]:
     grouped: dict[str, list[dict[str, str]]] = {
         "flag_shipped_default_off": [],
+        "flag_shipped_default_on": [],
         "substrate_shipped_flag_missing": [],
         "substrate_missing": [],
     }
diff --git a/evals/public_demo/results/v1_dev.json b/evals/public_demo/results/v1_dev.json
index 7b544c4d..01e42073 100644
--- a/evals/public_demo/results/v1_dev.json
+++ b/evals/public_demo/results/v1_dev.json
@@ -13,7 +13,7 @@
     {
       "case_id": "determinism_run_to_run_byte_equality",
       "details": {
-        "sha256": "f35ba6c28f008f7f80eef51fdecd36c951b7f38eea1d64dc163c5f45e5c86050"
+        "sha256": "b166949828921da9f45b19e3f7c2e6cf76a2073302244778dc2a4009bd4926f4"
       },
       "divergence": null,
       "passed": true
diff --git a/evals/public_demo/runner.py b/evals/public_demo/runner.py
index d60f0c17..b525e42b 100644
--- a/evals/public_demo/runner.py
+++ b/evals/public_demo/runner.py
@@ -69,6 +69,11 @@ _VOLATILE_KEYS: frozenset[str] = frozenset(
         "total_runtime_ms",  # wall-clock
         "json_path",  # adapter output paths (per-run temp dirs)
         "transient_corpus",  # learning-loop embeds its temp corpus path
+        # ``generated_at_revision`` advances every commit; the lane's
+        # invariant is "same code → same SHA," not "same HEAD → same
+        # SHA." Stripping this here keeps the pinned lane SHA stable
+        # across commits unless the underlying demos' content changes.
+        "generated_at_revision",
     }
 )
 
diff --git a/scripts/verify_lane_shas.py b/scripts/verify_lane_shas.py
index 022339cd..6d2269ce 100644
--- a/scripts/verify_lane_shas.py
+++ b/scripts/verify_lane_shas.py
@@ -49,7 +49,7 @@ PINNED_SHAS: dict[str, str] = {
     "domain_contract_validation": "f9c06cdeea8fb36a0d3c320007618c3afc92d67702ef31bd36ebd9ae9ced473f",
     "fabrication_control_summary": "01e1b6b711141f2b4a14551d7df3ea482d8d6dd7b364a25c509f4f8d08cda8a8",
     "demo_composition": "27d838241bf3ed9e15d0e918ec6d89a823494d7e17c2dab9777825af7188f20f",
-    "public_demo": "71090323943b3b09bf563d3d5217426922ab027f9a20bee3af4173435dda4c8c",
+    "public_demo": "4be6f47509435a24984713acfcebd88e61f4e1278096fa5dc88a09e8af2f87ba",
 }
 
 
diff --git a/tests/test_capability_reports.py b/tests/test_capability_reports.py
index 0d058073..d8215687 100644
--- a/tests/test_capability_reports.py
+++ b/tests/test_capability_reports.py
@@ -46,9 +46,19 @@ def test_chain_report_exposes_required_count_axes() -> None:
     assert len(philosophy_intents) >= 3
 
 
-def test_flag_report_tracks_default_off_flags_without_enabling_them() -> None:
+def test_flag_report_classification_matches_actual_defaults() -> None:
+    """Catalog classification must match DEFAULT_CONFIG's actual values.
+
+    A flag in ``flag_shipped_default_off`` MUST be False on DEFAULT_CONFIG;
+    a flag in ``flag_shipped_default_on`` MUST be True. Drift between the
+    catalog and the runtime default is itself a bug — catching it here
+    prevents the kind of silent re-classification that landed when
+    ``discourse_planner`` was flipped to True on 2026-05-21 without
+    updating the catalog.
+    """
     report = flag_report()
-    shipped = {row["flag"] for row in report["flag_shipped_default_off"]}
+    default_off = {row["flag"] for row in report["flag_shipped_default_off"]}
+    default_on = {row["flag"] for row in report["flag_shipped_default_on"]}
 
     for flag in (
         "realizer_grounded_authority",
@@ -56,17 +66,62 @@ def test_flag_report_tracks_default_off_flags_without_enabling_them() -> None:
         "transitive_surface",
         "gloss_aware_cause",
         "thread_anaphora",
-        "discourse_planner",
     ):
-        assert flag in shipped
-        assert getattr(DEFAULT_CONFIG, flag) is False
-    assert "stop_tokens" in shipped
+        assert flag in default_off, f"{flag!r} expected in default_off"
+        assert getattr(DEFAULT_CONFIG, flag) is False, (
+            f"{flag!r} catalog says default_off but DEFAULT_CONFIG has it True"
+        )
+
+    # discourse_planner shipped flag-off and was flipped to default-on
+    # on 2026-05-21 after cognition-eval byte-equality verification.
+    # The catalog state must match.
+    assert "discourse_planner" in default_on, (
+        "discourse_planner expected in default_on after the 2026-05-21 flip"
+    )
+    assert DEFAULT_CONFIG.discourse_planner is True
+
+    # stop_tokens is None-by-default (not a bool); kept in default_off
+    # because its absence is the "off" state.
+    assert "stop_tokens" in default_off
     assert DEFAULT_CONFIG.stop_tokens is None
+
     assert report["substrate_shipped_flag_missing"] == [
         {"flag": "compound_intent_dispatch", "adr": "ADR-0089-C2"}
     ]
 
 
+def test_flag_catalog_state_is_consistent_with_default_config() -> None:
+    """Cross-check every catalog entry against DEFAULT_CONFIG.
+
+    Catches any future drift: if someone changes a default in
+    ``core.config`` without updating the catalog state, this test
+    fails before the lane SHAs can shift.
+    """
+    report = flag_report()
+    for row in report["flag_shipped_default_off"]:
+        flag = row["flag"]
+        if not hasattr(DEFAULT_CONFIG, flag):
+            continue  # substrate flag without a runtime knob
+        value = getattr(DEFAULT_CONFIG, flag)
+        # ``stop_tokens`` is None-by-default (not a bool). All others
+        # should be ``False`` to honor the ``default_off`` classification.
+        if flag == "stop_tokens":
+            assert value is None
+        else:
+            assert value is False, (
+                f"catalog classifies {flag!r} as default_off but "
+                f"DEFAULT_CONFIG has it {value!r}"
+            )
+    for row in report["flag_shipped_default_on"]:
+        flag = row["flag"]
+        if not hasattr(DEFAULT_CONFIG, flag):
+            continue
+        assert getattr(DEFAULT_CONFIG, flag) is True, (
+            f"catalog classifies {flag!r} as default_on but "
+            f"DEFAULT_CONFIG has it {getattr(DEFAULT_CONFIG, flag)!r}"
+        )
+
+
 def test_ledger_status_is_predicate_derived() -> None:
     report = ledger_report()
     rows = {row["domain"]: row for row in report["domains"]}