fix: remote job stuck after SSH update completes (v0.5.4)

Treat agent_update_state as source of truth for active job; poll log thread no longer overwrites finished status.

backend/tests/test_agent_update.py

@@ -247,6 +247,30 @@ def test_clear_stale_running_remote_actions(db_session):
     assert host.remote_action["ok"] is False
 
 
+def test_get_remote_action_status_ignores_stale_running_payload(db_session):
+    from app.services.host_remote_actions import get_remote_action_status
+
+    host = Host(
+        hostname="done-linux",
+        os_family="linux",
+        product="ssh-monitor",
+        agent_update_state="success",
+        remote_action={
+            "status": "running",
+            "message": "Выполняется обновление… (лог с хоста)",
+            "output": "=== Script update completed successfully ===",
+            "ok": True,
+            "finished_at": "2026-07-08T02:21:11+00:00",
+        },
+    )
+    db_session.add(host)
+    db_session.commit()
+
+    status = get_remote_action_status(host)
+    assert status["active"] is False
+    assert status["agent_update_state"] == "success"
+
+
 def test_cancel_host_remote_job_api(jwt_headers, client, db_session):
     host = Host(
         hostname="cancel-me",