feat(auth): Optimize retry logic and add circuit breaker for team processing

- Reduce authorization retry attempts from 3 to 2 and optimize retry delays from exponential (3s, 5s, 8s, 12s) to fixed 3s backoff
- Implement team-level circuit breaker: skip member processing when 3+ consecutive 500 errors detected in same team
- Add consecutive 500 error tracking with atomic counter and reset on successful authorization
- Reduce ObtainAuthorizationCode timeout from 3 minutes to 90 seconds with updated error messages
- Optimize Codex API workspace selection retry: reduce max attempts from 5 to 3 with shorter backoff (2s, 4s instead of 3s, 5s, 8s, 12s)
- Reduce S2A OAuth submission retry delays from (4s, 6s) to (2s, 3s) for faster failure detection
- Optimize member stagger timing: reduce from 3s + idx*2s to 1s + idx*1s with reduced jitter (0-1s instead of 0-2s)
- Add early exit for exhausted retries in CompleteWithCodexAPI to prevent unnecessary outer retry attempts
- These changes improve responsiveness and reduce cascading failures during bulk team processing

backend/internal/api/team_process.go

@@ -730,6 +730,9 @@ func processSingleTeam(idx int, req TeamProcessRequest) (result TeamProcessResul
 	var s2aSuccessCount int32
 	var s2aFailCount int32
 
+	// Team 级别 500 错误熔断器：当同 Team 多个成员连续遇到 500 时快速失败
+	var consecutive500Fails int32
+
 	// 入库并发控制信号量
 	s2aSem := make(chan struct{}, req.ConcurrentS2A)
 
@@ -779,14 +782,23 @@ func processSingleTeam(idx int, req TeamProcessRequest) (result TeamProcessResul
 		var s2aSuccess bool
 		var lastError string
 
-		for attempt := 0; attempt < 3; attempt++ { // 最多重试2次
+		for attempt := 0; attempt < 2; attempt++ { // 最多重试1次
 			// 检查停止信号
 			if isStopped() {
 				return false
 			}
+			// 熔断检查：同 Team 已有 3+ 个成员连续 500 失败，快速跳过
+			if atomic.LoadInt32(&consecutive500Fails) >= 3 {
+				logger.Warning(fmt.Sprintf("%s 同 Team 已有多个成员连续 500 失败，跳过入库", memberLogPrefix), memberEmail, "team")
+				atomic.AddInt32(&s2aFailCount, 1)
+				memberMu.Lock()
+				result.Errors = append(result.Errors, fmt.Sprintf("成员 %d 入库跳过: 服务器持续 500 (熔断)", memberIdx+1))
+				memberMu.Unlock()
+				return false
+			}
 			if attempt > 0 {
-				// 重试前等待一段时间，避免密集请求（可被停止信号中断）
-				retryDelay := time.Duration(3+attempt*2) * time.Second
+				// 重试前短暂等待（可被停止信号中断）
+				retryDelay := time.Duration(3) * time.Second
 				logger.Warning(fmt.Sprintf("%s 入库重试 (第%d次, 等待 %ds)", memberLogPrefix, attempt+1, int(retryDelay.Seconds())), memberEmail, "team")
 				select {
 				case <-time.After(retryDelay):
@@ -847,8 +859,14 @@ func processSingleTeam(idx int, req TeamProcessRequest) (result TeamProcessResul
 			if err != nil {
 				lastError = fmt.Sprintf("浏览器授权失败: %v", err)
 				logger.Error(fmt.Sprintf("%s %s", memberLogPrefix, lastError), memberEmail, "team")
+				// 跟踪 500 错误用于熔断
+				if strings.Contains(err.Error(), "500") || strings.Contains(err.Error(), "重试已耗尽") {
+					atomic.AddInt32(&consecutive500Fails, 1)
+				}
 				continue
 			}
+			// 授权成功，重置 500 计数
+			atomic.StoreInt32(&consecutive500Fails, 0)
 
 			// 提交到 S2A
 			_, err = auth.SubmitS2AOAuth(
@@ -964,8 +982,8 @@ func processSingleTeam(idx int, req TeamProcessRequest) (result TeamProcessResul
 			s2aWg.Add(1)
 			go func(idx int, e, p string) {
 				defer s2aWg.Done()
-				// 基础 3 秒 + 成员索引 * 2 秒错开 + 0~2 秒随机抖动，避免同 Team 多成员同时选工作区
-				stagger := 3*time.Second + time.Duration(idx*2)*time.Second + time.Duration(rand.Intn(2000))*time.Millisecond
+				// 基础 1 秒 + 成员索引 * 1 秒错开 + 0~1 秒随机抖动
+				stagger := 1*time.Second + time.Duration(idx)*time.Second + time.Duration(rand.Intn(1000))*time.Millisecond
 				select {
 				case <-time.After(stagger):
 				case <-teamProcessState.stopCh: