fix(router): keep retry skips on the active order-fallback target

The retry-skip guard checks that some other deployment could still answer
before it excludes the one that just refused, so a single-deployment group
keeps the old retry-in-place behavior. It asked that question at the group's
minimum order, but the router picks the retry's deployment at the order the
request has already escalated to.

So a group with a primary at order 1 and a backup at order 2 answered "yes,
order 1 still has a candidate" while the retry was pinned to order 2, and the
exclusion left order 2 with nothing. The caller got a no-deployments error in
place of the provider's own 400.

The helper now takes the active target order and filters by it, which is the
same value async_get_healthy_deployments reads off the request.

litellm/router.py

@@ -402,6 +402,7 @@ def _stream_chunks_have_generated_content(chunks: Sequence[ModelResponseStream])
 _NO_SESSION_KWARGS: Final[Mapping[str, Mapping[str, object]]] = MappingProxyType({})
 _SESSION_ADAPTER: Final = TypeAdapter(Mapping[str, object])
 _EXCLUDED_DEPLOYMENT_IDS_ADAPTER: Final = TypeAdapter(tuple[str, ...])
+_TARGET_ORDER_ADAPTER: Final = TypeAdapter(int | None)
 
 
 def _with_router_resolved_session_model(session: object, model_name: str) -> Mapping[str, Mapping[str, object]]:
@@ -7464,6 +7465,7 @@ class Router:
         exception: Exception,
         already_skipped: object,
         healthy_deployments: list[dict],  # mutable-ok: matches the routing filters' list contract
+        target_order: object = None,
     ) -> tuple[str, ...]:
         failed_deployment_id: Final[str | None] = getattr(exception, "failed_deployment_id", None)
         status_code: Final = getattr(exception, "status_code", None)
@@ -7473,7 +7475,9 @@ class Router:
             return ()
         already_skipped_ids: Final = _EXCLUDED_DEPLOYMENT_IDS_ADAPTER.validate_python(already_skipped or ())
         skipped: Final = frozenset((*already_skipped_ids, failed_deployment_id))
-        same_order_candidates: Final = litellm.utils.get_order_filtered_deployments(healthy_deployments)
+        same_order_candidates: Final = litellm.utils.get_order_filtered_deployments(
+            healthy_deployments, target_order=_TARGET_ORDER_ADAPTER.validate_python(target_order)
+        )
         if not litellm.utils.get_excluded_filtered_deployments(same_order_candidates, excluded_deployment_ids=skipped):
             return ()
         verbose_router_logger.debug(
@@ -7580,6 +7584,7 @@ class Router:
                     exception=original_exception,
                     already_skipped=kwargs.get("_excluded_deployment_ids"),
                     healthy_deployments=_healthy_deployments,
+                    target_order=kwargs.get("_target_order"),
                 )
                 if skipped_deployment_ids:
                     kwargs["_excluded_deployment_ids"] = skipped_deployment_ids
@@ -7656,6 +7661,7 @@ class Router:
                         exception=e,
                         already_skipped=kwargs.get("_excluded_deployment_ids"),
                         healthy_deployments=_healthy_deployments,
+                        target_order=kwargs.get("_target_order"),
                     )
                     if retry_skipped_deployment_ids:
                         kwargs["_excluded_deployment_ids"] = retry_skipped_deployment_ids

tests/test_litellm/test_router.py

@@ -13261,6 +13261,36 @@ def test_router_deployment_ids_to_skip_on_retry(
     )
 
 
+@pytest.mark.parametrize(
+    "target_order,deployment_orders,expected",
+    [
+        (2, {"rejecting": 2, "sibling": 1}, ()),
+        (2, {"rejecting": 2, "sibling": 2}, ("rejecting",)),
+        (1, {"rejecting": 1, "sibling": 2}, ()),
+        (None, {"rejecting": 1, "sibling": 2}, ()),
+        (None, {"rejecting": 1, "sibling": 1}, ("rejecting",)),
+        (3, {"rejecting": 2, "sibling": 1}, ()),
+    ],
+)
+def test_router_deployment_ids_to_skip_on_retry_honors_order_fallback_target(
+    target_order, deployment_orders, expected
+):
+    exception = Exception("upstream refused this request")
+    exception.status_code = 400
+    exception.failed_deployment_id = "rejecting"
+    healthy_deployments = [
+        {"model_info": {"id": deployment_id}, "litellm_params": {"order": order}}
+        for deployment_id, order in deployment_orders.items()
+    ]
+
+    assert (
+        litellm.Router._deployment_ids_to_skip_on_retry(
+            exception, None, healthy_deployments, target_order=target_order
+        )
+        == expected
+    )
+
+
 def _make_failure_logging_obj():
     return LiteLLMLogging(
         model="gpt-5.6",