fix(swarm): don't paint healthy deckies as failed when a shard-sibling fails

docker compose up is partial-success-friendly — a build failure on one
service doesn't roll back the others. But the master was catching the
agent's 500 and tagging every decky in the shard as 'failed' with the
same error message. From the UI that looked like all three deckies died
even though two were live on the worker.

On dispatch exception, probe the agent's /status to learn which deckies
actually have running containers, and upsert per-decky state accordingly.
Only fall back to marking the whole shard failed if the status probe
itself is unreachable.

Enhance agent.executor.status() to include a 'runtime' map keyed by
decky name with per-service container state, so the master has something
concrete to consult.

tests/swarm/test_swarm_api.py

@@ -273,6 +273,68 @@ def test_deploy_rejects_missing_host_uuid(client: TestClient, stub_agent) -> Non
     assert "host_uuid" in resp.json()["detail"]
 
 
+def test_deploy_partial_failure_only_marks_actually_failed_decky(
+    client: TestClient, repo, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    """docker compose up is partial-success-friendly: one failed service
+    doesn't roll back the ones already up. The master must probe /status
+    after a dispatch exception so healthy deckies aren't painted red just
+    because a sibling in the same shard failed."""
+
+    class _PartialFailAgent:
+        def __init__(self, host=None, **_):
+            self._host = host or {}
+
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, *exc):
+            return None
+
+        async def deploy(self, config, **kw):
+            raise RuntimeError("Server error '500 Internal Server Error'")
+
+        async def status(self):
+            return {
+                "deployed": True,
+                "runtime": {
+                    "decky1": {"running": True, "services": {"ssh": "running"}},
+                    "decky2": {"running": True, "services": {"ssh": "running"}},
+                    "decky3": {"running": False, "services": {"ssh": "absent"}},
+                },
+            }
+
+    from decnet.web.router.swarm import api_deploy_swarm as deploy_mod
+    monkeypatch.setattr(deploy_mod, "AgentClient", _PartialFailAgent)
+
+    h1 = client.post(
+        "/swarm/enroll",
+        json={"name": "decktest", "address": "192.168.1.47", "agent_port": 8765},
+    ).json()
+
+    cfg = {
+        "mode": "swarm",
+        "interface": "eth0",
+        "subnet": "192.168.1.0/24",
+        "gateway": "192.168.1.1",
+        "deckies": [
+            _decky_dict("decky1", h1["host_uuid"], "192.168.1.2"),
+            _decky_dict("decky2", h1["host_uuid"], "192.168.1.3"),
+            _decky_dict("decky3", h1["host_uuid"], "192.168.1.4"),
+        ],
+    }
+    resp = client.post("/swarm/deploy", json={"config": cfg})
+    assert resp.status_code == 200
+    assert resp.json()["results"][0]["ok"] is False
+
+    shards = {s["decky_name"]: s for s in client.get("/swarm/deckies").json()}
+    assert shards["decky1"]["state"] == "running"
+    assert shards["decky1"]["last_error"] is None
+    assert shards["decky2"]["state"] == "running"
+    assert shards["decky3"]["state"] == "failed"
+    assert "500" in (shards["decky3"]["last_error"] or "")
+
+
 def test_deploy_rejects_non_swarm_mode(client: TestClient, stub_agent) -> None:
     cfg = {
         "mode": "unihost",