diff --git a/tests/test_sleep_engine.py b/tests/test_sleep_engine.py
index f3fb4e4..36cb88b 100644
--- a/tests/test_sleep_engine.py
+++ b/tests/test_sleep_engine.py
@@ -1333,6 +1333,8 @@ class TestVerifierDiscipline(unittest.TestCase):
         """Gate must reject an edit that partially degrades held-out (1.0→0.5),
         not just all-or-nothing collapses. Train improves (0.0→1.0) which makes
         the regression easy to miss — the gate catches it anyway."""
+        from skillopt_sleep.replay import aggregate_scores, replay_batch
+
         be = MockFractionalDegradationBackend()
         train = TaskRecord(id="t3", project="/p", intent="train", reference="ABC",
                            reference_kind="exact", tags=["rule:__reward_hacking__"], split="train")
@@ -1342,6 +1344,11 @@ class TestVerifierDiscipline(unittest.TestCase):
                           reference_kind="exact", tags=["rule:real"], split="val")
         tasks = [train, val1, val2]
 
+        candidate_pairs = replay_batch(be, [val1, val2], be.HACK_TEXT, "")
+        candidate_hard, _candidate_soft = aggregate_scores(candidate_pairs)
+        self.assertEqual([result.hard for _task, result in candidate_pairs], [1.0, 0.0])
+        self.assertEqual(candidate_hard, 0.5)
+
         res = consolidate(be, tasks, "", "", edit_budget=4, gate_metric="hard", night=1)
 
         self.assertFalse(res.accepted)
@@ -1356,7 +1363,6 @@ class TestVerifierDiscipline(unittest.TestCase):
         self.assertIn("definitive answer", res.rejected_edits[0].content)
 
 
-
 class TestDiagnosticsRedaction(unittest.TestCase):
     """diagnostics.json surfaces backend stderr / optimizer replies / task
     responses for debugging — but those can carry credentials (e.g. a codex 401