Correct the one-controller claim: the drone takes over, it does not refuse

Tested on the aircraft. A second ARSDK handshake is accepted and telemetry is
redirected to it; the first session's frames stop while it still reports
connected = True. So the claim inherited from pyparrot's error text, which had
reached our error messages, tool descriptions, simulator behaviour and a test
name, was wrong in the most misleading direction: a refusal would be loud, and
this is silent.

The simulator now models the takeover by default; refusal stays available
because a client must handle a non-zero status anyway.

tests/test_arsdk_session.py

@@ -119,7 +119,14 @@ async def test_handshake_describes_this_controller(sim, session):
     assert request["d2c_port"] == session.d2c_port != 0
 
 
-async def test_second_controller_is_refused(sim, session):
+async def test_second_controller_takes_the_link_over(sim, session):
+    """The aircraft accepts a newcomer and stops talking to whoever had it.
+
+    Measured on the real drone 2026-10-02: the first session's frame count
+    froze while the second received, and the first went on reporting
+    connected = True. That silent failure is why a client should judge a link
+    by telemetry freshness rather than by its connected flag.
+    """
     other = DroneSession(
         "127.0.0.1",
         discovery_port=sim.discovery_port,
@@ -128,9 +135,13 @@ async def test_second_controller_is_refused(sim, session):
         encoder=fake_encode,
         decoder=fake_decode,
     )
-    with pytest.raises(HandshakeError, match="one controller"):
-        await other.connect()
-    assert not other.connected
+    await other.connect()
+    try:
+        assert other.connected
+        # The loser keeps its socket and its optimism; only the data stops.
+        assert session.connected
+    finally:
+        await other.disconnect()
 
 
 async def test_unreachable_address_fails_fast_rather_than_hanging():