fix repeated successful evidence call loops

src/clean_agent_preview.py

@@ -5120,12 +5120,21 @@ async def stream_preview(*, endpoint_url, model, messages, headers, turn_contrac
                             calls += 1
                             duplicate_count = successful_duplicate_counts.get(call_signature, 0) + 1
                             successful_duplicate_counts[call_signature] = duplicate_count
-                            if evidence_tool_keeps_distinct_requests_available(name):
+                            if (
+                                evidence_tool_keeps_distinct_requests_available(name)
+                                and duplicate_count < 2
+                            ):
                                 suppression = (
                                     'rejected only for this exact request; the tool remains '
                                     'available with different arguments'
                                 )
                             elif duplicate_count >= 2:
+                                # A first duplicate leaves evidence tools available so a
+                                # corrected page/range/URL/query can execute immediately.
+                                # Repeating that exact successful call after the reminder
+                                # proves the model is not taking that route; continuing to
+                                # advertise it produces no-op loops and starves completion.
+                                terminal_suppression_violation = True
                                 permanently_suppressed_tools.add(canonical(name))
                                 suppression = 'disabled for the rest of this turn'
                             else:

tests/test_clean_agent_preview.py

@@ -5251,7 +5251,10 @@ async def test_native_stream_permanently_withholds_tool_when_model_ignores_dupli
         and "already returned evidence" in event.get("output", "")
     ]
     assert len(duplicate_errors) == 2
-    assert 'tools' not in requests[2]
+    # The first exact duplicate leaves the evidence tool available so a
+    # changed request can run.  Ignoring that correction once more suppresses
+    # the looping tool and forces a completion-only turn.
+    assert any(item['function']['name'] == 'inspect_media' for item in requests[2]['tools'])
     assert 'tools' not in requests[3]