[eric] backend: fast-fail is_running() to kill the ~18s cold-start event-loop freeze

- faulthandler on the signed cold build caught the asyncio loop frozen in
  socket.create_connection inside is_running() (process.py): a synchronous
  httpx.get to "localhost:20128" called ~5x on the boot path before 9Router is up
- on Windows a dead-port connect to "localhost" stalls ~7s each (tries ::1 first,
  loopback refusal is slow), freezing the loop ~18s so uvicorn could not answer
  the health probe -> cold backend-http-ready was ~23s
- fix: probe 127.0.0.1 with a 0.3s TCP timeout first (measured 306ms vs ~7s), only
  HTTP-confirm when the port is open; 9Router binds 0.0.0.0 so reachability is
  unchanged. drop the faulthandler diagnostic from main.py

backend/apps/nine_router/process.py

@@ -16,6 +16,7 @@ import logging
 import os
 import secrets
 import shutil
+import socket
 import subprocess
 import tempfile
 import time
@@ -74,13 +75,29 @@ _is_running_last_ok: float = 0.0
 
 
 def is_running() -> bool:
-    """Check if 9Router is running."""
+    """Check if 9Router is running.
+
+    Fast-fail when down. is_running() is called ~5x on the cold boot path (the
+    settings key-sync sequence + ensure_running) BEFORE 9Router is up. The old
+    body did a synchronous httpx.get to "localhost:20128"; on Windows a dead-port
+    connect to "localhost" stalls multiple seconds (it tries ::1 first and the
+    loopback refusal is slow), so those probes froze the asyncio event loop ~18s
+    and dominated cold startup (faulthandler caught the loop stuck in
+    socket.create_connection here). Fix: probe 127.0.0.1 with a 0.3s TCP timeout
+    first; a down 9Router is detected in <~0.3s instead of ~7s. Only when the
+    port is open do we do the HTTP confirm. 9Router binds 0.0.0.0 (the warm app
+    reaches it via 127.0.0.1 today), so this changes timing, not reachability."""
     global _is_running_last_ok
     now = time.monotonic()
     if now - _is_running_last_ok < _IS_RUNNING_TTL:
         return True
     try:
-        r = httpx.get(f"{NINE_ROUTER_V1}/models", timeout=2.0)
+        with socket.create_connection(("127.0.0.1", NINE_ROUTER_PORT), timeout=0.3):
+            pass
+    except OSError:
+        return False
+    try:
+        r = httpx.get(f"http://127.0.0.1:{NINE_ROUTER_PORT}/v1/models", timeout=2.0)
         if r.status_code == 200:
             _is_running_last_ok = now
             return True

backend/main.py

@@ -18,18 +18,6 @@ if not _backend_logger.handlers:
 
 logger = logging.getLogger(__name__)
 
-# [perf][diagnostic] Cold-start stall hunt: dump every thread's stack every 7s to
-# a temp file. During the cold ~13s event-loop stall a dump lands inside the
-# frozen window and names the exact synchronous call the loop is stuck in. Temp
-# file only, best-effort; remove once the stall is diagnosed.
-try:
-    import faulthandler as _faulthandler
-    import tempfile as _tempfile
-    _fh_diag = open(os.path.join(_tempfile.gettempdir(), "openswarm-faulthandler.log"), "w")
-    _faulthandler.dump_traceback_later(7, repeat=True, file=_fh_diag)
-except Exception:
-    pass
-
 from fastapi.responses import JSONResponse, HTMLResponse
 from fastapi import Request