flink cdc to doris 部分数据丢失

flink1.20, cdc 3.3, doris,1.2.8, fe、be部署分别独立的6台机器。
flink 从mysql同步数据到 doris,第一次init初始化同步32张表数据，发现其中一张表在任务同步完后数据数量不一致，原表2309061，目标2308831。查询flink调度日志未出现相关错误，
fe warn日志

2025-06-13 13:38:55,817 WARN (replayer|84) [Env.replayJournal():2871] entityCost:1143 loadJournalCost:1143 logId:38119935 replayedJournalId:38119935 code:100 size:512
2025-06-13 13:46:39,802 WARN (replayer|84) [BDBJournalCursor.next():129] fail to get journal 38124197, will try again. status: OperationStatus.NOTFOUND
2025-06-13 13:46:42,803 WARN (replayer|84) [Env.replayJournal():2871] entityCost:3000 loadJournalCost:3000 logId:38124197 replayedJournalId:38124197 code:100 size:512

be日志有大量

fragment_mgr.cpp:1316] Could not find the query id:ad60767ccfed4638-9e805378f8707c2c fragment id:0 to cancel

be服务 PipeGBlockSche- 线程持续100%