test(audit): fix critical test suite deadlocks and write exhaustive architectural report

- Fix 'Triple Bingo' history synchronization explosion during streaming

- Implement stateless event buffering in ApiHookClient to prevent dropped events

- Ensure 'tool_execution' events emit consistently across all LLM providers

- Add hard timeouts to all background thread wait() conditions

- Add thorough teardown cleanup to conftest.py's reset_ai_client fixture

- Write highly detailed report_gemini.md exposing asyncio lifecycle flaws

tests/conftest.py

@@ -56,6 +56,28 @@ class VerificationLogger:
    f.write(f"{status} {self.test_name} ({result_msg})\n\n")
   print(f"[FINAL] {self.test_name}: {status} - {result_msg}")
 
+@pytest.fixture(autouse=True)
+def reset_ai_client() -> Generator[None, None, None]:
+ """
+ Autouse fixture that resets the ai_client global state before each test.
+ This is critical for preventing state pollution between tests.
+ """
+ import ai_client
+ import mcp_client
+ ai_client.reset_session()
+ # Reset callbacks to None or default to ensure no carry-over
+ ai_client.confirm_and_run_callback = None
+ ai_client.comms_log_callback = None
+ ai_client.tool_log_callback = None
+ # Clear all event listeners
+ ai_client.events.clear()
+ # Reset provider/model to defaults
+ ai_client.set_provider("gemini", "gemini-2.5-flash-lite")
+ # Reset MCP client state
+ mcp_client.configure([], [])
+ yield
+ ai_client.reset_session()
+
 @pytest.fixture
 def vlogger(request) -> VerificationLogger:
  """Fixture to provide a VerificationLogger instance to a test."""
@@ -109,8 +131,8 @@ def mock_app() -> Generator[App, None, None]:
   app = App()
   yield app
   if hasattr(app, 'controller'):
-   app.controller.stop_services()
-  if hasattr(app, 'shutdown'):
+   app.controller.shutdown()
+  elif hasattr(app, 'shutdown'):
    app.shutdown()
 
 @pytest.fixture
@@ -142,7 +164,7 @@ def app_instance() -> Generator[App, None, None]:
   yield app
   # Cleanup: Ensure background threads and asyncio loop are stopped
   if hasattr(app, 'controller'):
-   app.controller.stop_services()
+   app.controller.shutdown()
 
   if hasattr(app, 'shutdown'):
    app.shutdown()
@@ -209,10 +231,13 @@ def live_gui() -> Generator[tuple[subprocess.Popen, str], None, None]:
  
  # Check if already running (shouldn't be)
  try:
-  resp = requests.get("http://127.0.0.1:8999/status", timeout=0.1)
-  already_up = resp.status_code == 200
- except: already_up = False
- diag.log_state("Hook Server Port 8999", "Down", "UP" if already_up else "Down")
+  resp = requests.get("http://127.0.0.1:8999/status", timeout=0.5)
+  if resp.status_code == 200:
+   print("[Fixture] WARNING: Hook Server already up on port 8999. Test state might be polluted.")
+   # Optionally try to reset it
+   try: requests.post("http://127.0.0.1:8999/api/gui", json={"action": "click", "item": "btn_reset"}, timeout=1)
+   except: pass
+ except: pass
 
  print(f"\n[Fixture] Starting {gui_script} --enable-test-hooks in {temp_workspace}...")
  os.makedirs("logs", exist_ok=True)

tests/test_extended_sims.py

@@ -52,7 +52,6 @@ def test_tools_sim_live(live_gui: Any) -> None:
  sim.run() # Ensure history is updated via the async queue
  time.sleep(2)
  sim.teardown()
-
 @pytest.mark.integration
 def test_execution_sim_live(live_gui: Any) -> None:
  """Run the Execution & Modals simulation against a live GUI."""
@@ -60,7 +59,11 @@ def test_execution_sim_live(live_gui: Any) -> None:
  assert client.wait_for_server(timeout=10)
  sim = ExecutionSimulation(client)
  sim.setup("LiveExecutionSim")
+ # Enable manual approval to test modals
+ client.set_value('manual_approve', True)
  client.set_value('current_provider', 'gemini_cli')
  client.set_value('gcli_path', f'"{sys.executable}" "{os.path.abspath("tests/mock_gemini_cli.py")}"')
  sim.run()
+ time.sleep(2)
  sim.teardown()
+

tests/test_gemini_cli_edge_cases.py

@@ -56,7 +56,8 @@ def test_gemini_cli_parameter_resilience(live_gui: Any) -> None:
     """
  client = ApiHookClient("http://127.0.0.1:8999")
  client.click("btn_reset")
- time.sleep(1.5)
+ time.sleep(1.0)
+
  client.set_value("auto_add_history", True)
  client.set_value("manual_approve", True)
  client.select_list_item("proj_files", "manual_slop")
@@ -130,7 +131,8 @@ def test_gemini_cli_loop_termination(live_gui: Any) -> None:
     """
  client = ApiHookClient("http://127.0.0.1:8999")
  client.click("btn_reset")
- time.sleep(1.5)
+ time.sleep(1.0)
+
  client.set_value("auto_add_history", True)
  client.set_value("manual_approve", True)
  client.select_list_item("proj_files", "manual_slop")

tests/test_gemini_cli_integration.py

@@ -13,7 +13,8 @@ def test_gemini_cli_full_integration(live_gui: Any) -> None:
  client = ApiHookClient("http://127.0.0.1:8999")
  # 0. Reset session and enable history
  client.click("btn_reset")
- time.sleep(1.5)
+ time.sleep(1.0)
+
  client.set_value("auto_add_history", True)
  client.set_value("manual_approve", True)
  # Switch to manual_slop project explicitly
@@ -80,7 +81,8 @@ def test_gemini_cli_rejection_and_history(live_gui: Any) -> None:
  client = ApiHookClient("http://127.0.0.1:8999")
  # 0. Reset session
  client.click("btn_reset")
- time.sleep(1.5)
+ time.sleep(1.0)
+
  client.set_value("auto_add_history", True)
  client.set_value("manual_approve", True)
  client.select_list_item("proj_files", "manual_slop")

tests/test_visual_mma.py

@@ -68,3 +68,11 @@ def test_visual_mma_components(live_gui):
   assert tickets[1]['status'] == "running"
 
  print("Visual MMA component verification PASSED.")
+ 
+ # Clean up the pending modal to prevent polluting subsequent tests
+ print("Cleaning up pending MMA modal...")
+ client.post_gui({
+  "action": "click",
+  "item": "btn_approve_mma_step"
+ })
+ time.sleep(0.5)