fix: make error passthrough effective for non-failover upstream errors

backend/internal/handler/gateway_handler.go

@@ -135,6 +135,11 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 	// Track if we've started streaming (for error handling)
 	streamStarted := false
 
+	// 绑定错误透传服务，允许 service 层在非 failover 错误场景复用规则。
+	if h.errorPassthroughService != nil {
+		service.BindErrorPassthroughService(c, h.errorPassthroughService)
+	}
+
 	// 获取订阅信息（可能为nil）- 提前获取用于后续检查
 	subscription, _ := middleware2.GetSubscriptionFromContext(c)
 

backend/internal/handler/gemini_v1beta_handler.go

@@ -207,6 +207,9 @@ func (h *GatewayHandler) GeminiV1BetaModels(c *gin.Context) {
 
 	// 1) user concurrency slot
 	streamStarted := false
+	if h.errorPassthroughService != nil {
+		service.BindErrorPassthroughService(c, h.errorPassthroughService)
+	}
 	userReleaseFunc, err := geminiConcurrency.AcquireUserSlotWithWait(c, authSubject.UserID, authSubject.Concurrency, stream, &streamStarted)
 	if err != nil {
 		googleError(c, http.StatusTooManyRequests, err.Error())

backend/internal/handler/openai_gateway_handler.go

@@ -149,6 +149,11 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 	// Track if we've started streaming (for error handling)
 	streamStarted := false
 
+	// 绑定错误透传服务，允许 service 层在非 failover 错误场景复用规则。
+	if h.errorPassthroughService != nil {
+		service.BindErrorPassthroughService(c, h.errorPassthroughService)
+	}
+
 	// Get subscription info (may be nil)
 	subscription, _ := middleware2.GetSubscriptionFromContext(c)