[CI] Skip exited descendants when finding the worker in the FT e2e test (#61648)

Signed-off-by: Oxana Korzh <okorzh@amd.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>

tests/v1/fault_tolerance/test_fault_tolerance_e2e.py

@@ -190,11 +190,13 @@ def _apply_ft(server, instruction: str, params: dict | None = None) -> dict:
 
 def _kill_worker_process(server) -> None:
     """SIGKILL only the worker proc, leaving EngineCore and API server alive."""
-    workers = [
-        p
-        for p in psutil.Process(server.proc.pid).children(recursive=True)
-        if "Worker" in " ".join(p.cmdline())
-    ]
+    workers = []
+    for p in psutil.Process(server.proc.pid).children(recursive=True):
+        # Skip descendants that already exited: cmdline() raises ZombieProcess
+        # if unreaped, NoSuchProcess if reaped. A dead worker still fails below.
+        with contextlib.suppress(psutil.ZombieProcess, psutil.NoSuchProcess):
+            if "Worker" in " ".join(p.cmdline()):
+                workers.append(p)
     assert len(workers) == 1, f"expected 1 worker proc, found: {workers}"
     workers[0].kill()