fix(maintenance): a host that came back late stops reading as failed once it is back (#6105)

Docker02's Sunday job updated and rebooted, then failed only because
portal_app was short of its replica past the 30-minute settle. Its
maintenance_failed figure stayed 1, and the alert kept firing, until a
later job succeeds. A job failed for not coming back into service is
checked again each minute and reads 0 once the host is back; a failed
or timed-out update still reads 1 until a job succeeds.

Also: compose.deploy.yml's syslog comment still said devices are told
apart by address under Swarm (stale since #6096).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/integration/test_host_maintenance_cadence.py

@@ -134,3 +134,45 @@ def test_hosts_go_one_at_a_time_and_a_failure_stops_the_run():
         assert flags[seed["hosts"]["b"].name] == 1.0 and flags[seed["hosts"]["a"].name] == 0.0
 
     asyncio.run(_go())
+
+
+def test_a_host_that_came_back_late_reads_failed_only_until_it_is_back():
+    """#6105: a job whose update went through, failed only for not coming back
+    into service within the limit, stops flagging the host once it is back."""
+    app = _app()
+
+    async def _go():
+        from sqlalchemy import text
+        seed = await _seed(app)
+        cadence, models = _ha("cadence"), _ha("models")
+        async with app.db_sessionmaker() as s:
+            async with s.begin():
+                plan = await s.get(models.HostMaintenancePlan, seed["plan_id"])
+                await cadence.start_run(s, plan, datetime.now(timezone.utc), manual=True)
+        await _tick(app)
+        ta, a = seed["tokens"]["a"], seed["hosts"]["a"]
+        (step,) = await _post(app, ta)
+        await _post(app, ta, [{"kind": "maintenance", "id": step["id"], "action": "update",
+                               "state": "done", "ok": True, "packages": 2, "reboot_required": False}])
+
+        async def _sql(sql):
+            async with app.db_sessionmaker() as s:
+                async with s.begin():
+                    await s.execute(text(sql), {"h": a.id, "p": seed["plan_id"]})
+
+        # The agent goes quiet after its update, and stays so past the limit.
+        await _sql("UPDATE host_agent_registrations SET last_seen_at = now() - interval '1 hour' "
+                   "WHERE host_id = :h")
+        await _tick(app)
+        await _sql("UPDATE host_maintenance_runs SET settle_since = now() - interval '31 minutes' "
+                   "WHERE plan_id = :p")
+        flags = await _tick(app)
+        state, _, detail = await _run(app, seed["plan_id"])
+        assert state == "failed" and "not back in service within 30 min" in detail
+        assert flags[a.name] == 1.0
+
+        # Posting again, it is back in service: nothing is left failed.
+        await _post(app, ta)
+        assert (await _tick(app))[a.name] == 0.0
+
+    asyncio.run(_go())