Test Failure Analysis

Summary: A single timing-sensitive test failed on the Python 3.10 Ubuntu runner due to CI scheduling jitter. The PR changes (documentation only) are unrelated to the failure.

Root Cause: `tests/server/middleware/test_ping.py::TestPingLoop::test_ping_loop_sends_pings_at_interval` is a flaky test that relies on precise sub-second async scheduling. The test runs a ping loop with a 50ms interval for 350ms and asserts at least 2 pings are sent. On a loaded CI runner, `anyio.sleep(0.05)` can be delayed significantly by the scheduler, resulting in only 1 ping completing before the 350ms deadline.

The ping loop implementation (`ping.py:63-70`) sleeps before sending the first ping:
```python
while True:
await anyio.sleep(self.interval_ms / 1000) # sleep first
await session.send_ping() # then ping
```

With 350ms window ÷ 50ms interval = ~7x ratio, this should theoretically yield ~6 pings — but under scheduler pressure, even a single sleep can exhaust the window.

Suggested Solution: Increase the observation window in `tests/server/middleware/test_ping.py:172` relative to the interval to give more headroom for CI scheduling jitter. For example, change the window from `0.35` to `0.5` seconds (10x the interval), which provides enough buffer for typical CI delays while still verifying the looping behavior:

```python

Before

with anyio.move_on_after(0.35): # 7x the 50ms interval
await middleware._ping_loop(mock_session, session_id)
assert mock_session.send_ping.call_count >= 2

After

with anyio.move_on_after(0.5): # 10x the 50ms interval
await middleware._ping_loop(mock_session, session_id)
assert mock_session.send_ping.call_count >= 2
```

with anyio.move_on_after(0.35):
    await middleware._ping_loop(mock_session, session_id)

# Should have sent at least 2 pings in 350ms with 50ms interval
assert mock_session.send_ping.call_count >= 2