fix(providers): allow retry and fallback on stream stalled timeout

When a stream stalls mid-response, both the retry layer and
FallbackProvider blocked recovery because content had already been
emitted via on_content_delta. This left users with truncated replies
and no automatic recovery.

For error_kind="timeout" specifically:
- _run_with_retry now suppresses delta callbacks and retries the same
  model instead of returning immediately
- FallbackProvider now allows failover to a different model with
  delta callbacks suppressed

Non-timeout errors retain the original "skip retry/failover after
streamed content" behavior to avoid duplicate output.

nanobot/providers/base.py

@@ -827,10 +827,22 @@ class LLMProvider(ABC):
                 return response
             last_response = response
             if should_retry_guard is not None and not should_retry_guard():
-                logger.warning(
-                    "LLM stream failed after content was emitted; skipping retry"
-                )
-                return response
+                is_timeout = (response.error_kind or "").lower() == "timeout"
+                if is_timeout:
+                    logger.warning(
+                        "LLM stream stalled after content was emitted; "
+                        "suppressing delta callbacks and retrying"
+                    )
+                    kw.setdefault("on_content_delta", None)
+                    kw["on_content_delta"] = None
+                    kw["on_thinking_delta"] = None
+                    kw["on_tool_call_delta"] = None
+                    should_retry_guard = None
+                else:
+                    logger.warning(
+                        "LLM stream failed after content was emitted; skipping retry"
+                    )
+                    return response
             error_key = ((response.content or "").strip().lower() or None)
             if error_key and error_key == last_error_key:
                 identical_error_count += 1

nanobot/providers/fallback_provider.py

@@ -149,10 +149,20 @@ class FallbackProvider(LLMProvider):
                 return response
 
             if has_streamed is not None and has_streamed[0]:
-                logger.warning(
-                    "Primary model error but content already streamed; skipping failover"
-                )
-                return response
+                is_timeout = (response.error_kind or "").lower() == "timeout"
+                if is_timeout:
+                    logger.warning(
+                        "Primary model '{}' stream stalled after content was emitted; "
+                        "attempting failover anyway",
+                        primary_model,
+                    )
+                    has_streamed[0] = False
+                    kwargs["on_content_delta"] = None
+                else:
+                    logger.warning(
+                        "Primary model error but content already streamed; skipping failover"
+                    )
+                    return response
 
             if not self._should_fallback(response):
                 logger.warning(

tests/agent/test_runner_fallback.py

@@ -287,7 +287,7 @@ class TestFallbackOnPrimaryError:
 
 class TestNoFallbackWhenContentStreamed:
     @pytest.mark.asyncio
-    async def test(self) -> None:
+    async def test_non_timeout_error_skips_failover(self) -> None:
         primary = _FakeProvider("primary", _error_response())
         factory = MagicMock()
         fb = FallbackProvider(
@@ -303,12 +303,40 @@ class TestNoFallbackWhenContentStreamed:
             messages=[{"role": "user", "content": "hi"}],
             on_content_delta=_delta,
         )
-        # Primary returns error but content was "streamed" (FakeProvider calls delta)
-        # so failover should be skipped
         assert result.finish_reason == "error"
         factory.assert_not_called()
 
 
+class TestFallbackOnStreamStalledAfterContent:
+    @pytest.mark.asyncio
+    async def test_timeout_with_streamed_content_falls_back(self) -> None:
+        primary = _FakeProvider(
+            "primary",
+            _make_response("stream stalled", finish_reason="error", error_kind="timeout"),
+        )
+        fallback = _FakeProvider("fallback", _make_response("fallback ok"))
+        factory = MagicMock(return_value=fallback)
+        fb = FallbackProvider(
+            primary=primary,
+            fallback_presets=[_fallback("fallback-a")],
+            provider_factory=factory,
+        )
+
+        streamed: list[str] = []
+
+        async def _delta(text: str) -> None:
+            streamed.append(text)
+
+        result = await fb.chat_stream(
+            messages=[{"role": "user", "content": "hi"}],
+            on_content_delta=_delta,
+        )
+        assert result.finish_reason == "stop"
+        assert result.content == "fallback ok"
+        factory.assert_called_once_with(_fallback("fallback-a"))
+        assert "stream stalled" in streamed
+
+
 class TestFailoverOnTransientError:
     @pytest.mark.asyncio
     async def test_rate_limit(self) -> None:

tests/providers/test_provider_retry.py

@@ -163,6 +163,42 @@ async def test_chat_stream_with_retry_does_not_retry_after_emitting_content(monk
     assert delays == []
 
 
+@pytest.mark.asyncio
+async def test_chat_stream_with_retry_retries_timeout_after_emitting_content(monkeypatch) -> None:
+    first = LLMResponse(
+        content="Error calling LLM: stream stalled for more than 30 seconds",
+        finish_reason="error",
+        error_kind="timeout",
+    )
+    first._test_stream_delta = "partial"  # type: ignore[attr-defined]
+    provider = ScriptedProvider([
+        first,
+        LLMResponse(content="full retry response"),
+    ])
+    deltas: list[str] = []
+    delays: list[int] = []
+
+    async def _fake_sleep(delay: int) -> None:
+        delays.append(delay)
+
+    async def _on_delta(delta: str) -> None:
+        deltas.append(delta)
+
+    monkeypatch.setattr("nanobot.providers.base.asyncio.sleep", _fake_sleep)
+
+    response = await provider.chat_stream_with_retry(
+        messages=[{"role": "user", "content": "hello"}],
+        on_content_delta=_on_delta,
+    )
+
+    assert response.content == "full retry response"
+    assert response.finish_reason == "stop"
+    assert provider.calls == 2
+    assert deltas == ["partial"]
+    assert delays == [1]
+    assert provider.last_kwargs.get("on_content_delta") is None
+
+
 @pytest.mark.asyncio
 async def test_chat_with_retry_uses_provider_generation_defaults() -> None:
     """When callers omit generation params, provider.generation defaults are used."""