fix: widen thread-leak fixture grace period to match stop()'s timeout

test/conftest.py

@@ -21,18 +21,28 @@ def feeder():
 
 @pytest.fixture(autouse=True)
 def _assert_no_j1939_thread_leak():
-    """Fail any test that leaves a j1939.* background thread alive."""
+    """Fail any test that leaves a j1939.* background thread alive.
+
+    The poll window here (3.5s) is deliberately kept slightly above
+    ElectronicControlUnit.stop()'s own default dispatch_join_timeout (3.0s):
+    stop() only logs a warning (rather than failing) if the dispatch thread
+    doesn't exit within that timeout, e.g. while a slow subscriber callback
+    unwinds. If this fixture's window were shorter than stop()'s own
+    tolerance, a thread that stop() itself considers "still fine, just slow"
+    could trip a false-positive leak failure here. See #81.
+    """
     before = {t.ident for t in threading.enumerate()
               if t.name.startswith('j1939.')}
     yield
-    # Give freshly-stopped threads a brief moment to actually exit.
+    # Give freshly-stopped threads a chance to actually exit.
     import time
-    for _ in range(20):
+    deadline = time.monotonic() + 3.5
+    while True:
         leaked = [t for t in threading.enumerate()
                   if t.name.startswith('j1939.')
                   and t.ident not in before
                   and t.is_alive()]
-        if not leaked:
+        if not leaked or time.monotonic() >= deadline:
             break
         time.sleep(0.01)
     assert not leaked, (