Stale-feed watchdog treats 30s of user-hub silence as dead and force-reconnects a healthy suite

[TexasCoding]

Summary

HealthMonitoringMixin._stale_feed_watchdog_loop uses one threshold (stale_feed_seconds = 30) for both hubs. The user hub only emits order/position/account events. While flat (or overnight with no fills), 30s of user-hub silence is normal.

The watchdog then logs User feed stale; forcing reconnect even when health_score is 100, calls force_health_reconnect() → disconnect() + connect(), and can crash the process:

RuntimeError: Cannot connect while not disconnected

Related to the watchdog added for #97 (market tick freeze). Market-hub silence is a real signal. User-hub silence is not.

Evidence (project-x-py 4.1.1, MNQ+MES, no position)

User feed stale; forcing reconnect  hub=user seconds_since_message=30.00 threshold_seconds=30.0
Forcing reconnection due to poor health: 100.0
Health-based reconnection failed
...
RuntimeError: Cannot connect while not disconnected

Process then died (SIGTERM) ~8s later as the suite disconnected.

Suggested fix

• Do not treat user hub silence as stale, or use a much larger timeout (minutes, not 30s).
• Keep market hub stale detection (quotes/depth/ticks).
• force_health_reconnect should not call connect() while the transport is still connecting / not fully disconnected (Cannot connect while not disconnected).
• Do not force-reconnect when health_score is 100.

Workaround

After TradingSuite.create(), wrap realtime._hub_silence_age so hub == "user" returns None.