flinksql诡异问题，诡异很诡异！

flink版本是1.17

框架是flink on hive 执行sql任务：

一个sql，聚合过程的过程，包含多个count+sum过程；
但是，跑出来的结果值有时候为0，有时候正常！！
去掉其中部分count或者sum数据时，有时正常有时异常！
完整sql：
create table test.20230827_tmp9 as
select
countryid,
r_type,
c_type,
c_version,
count(distinct uid) user_cnt,
count(distinct did) dev_cnt,
count(distinct concat(did,cn)) channel_cnt,
count(distinct hc) hc_cnt,
count(1) all_cnt,
count(case when r not in (-1,-2,21,20084,25404,25409,25416,25451,25544,25546,26109,10053,10057,10121,10122,10125,10127) then 1 end) all_cnt_qu,
count(case when r = 0 then 1 end) success_cnt,
count(case when r = 0 and ft between 0 and 30000 then 1 end) success_cnt2,
sum(case when r = 0 and ft between 0 and 30000 then flowtbt else 0 end) flowtbt,
sum(case when r = 0 and ft between 0 and 30000 then flowtht else 0 end) flowtht,
sum(case when r = 0 and ft between 0 and 30000 then flowtot else 0 end) flowtot,
sum(case when r = 0 and ft between 0 and 30000 then decodet else 0 end) decodet,
sum(case when r = 0 and ft between 0 and 30000 then ft else 0 end) flowt,
sum(case when r = 0 and ft between 0 and 30000 then viewt else 0 end) viewt
from temp_table
where pdt = '2023-08-17'
group by
country_id,
r_type,
c_type,
c_version;
此时执行结果异常，从all_cnt字段开始，所有结果都是0；

正常sql（异常sql中去掉c_version字段到all_cnt之间几个count过程）：
create table test.20230827_tmp9 as
select
countryid,
r_type,
c_type,
c_version,
count(1) all_cnt,
count(case when r not in (-1,-2,21,20084,25404,25409,25416,25451,25544,25546,26109,10053,10057,10121,10122,10125,10127) then 1 end) all_cnt_qu,
count(case when r = 0 then 1 end) success_cnt,
count(case when r = 0 and ft between 0 and 30000 then 1 end) success_cnt2,
sum(case when r = 0 and ft between 0 and 30000 then flowtbt else 0 end) flowtbt,
sum(case when r = 0 and ft between 0 and 30000 then flowtht else 0 end) flowtht,
sum(case when r = 0 and ft between 0 and 30000 then flowtot else 0 end) flowtot,
sum(case when r = 0 and ft between 0 and 30000 then decodet else 0 end) decodet,
sum(case when r = 0 and ft between 0 and 30000 then ft else 0 end) flowt,
sum(case when r = 0 and ft between 0 and 30000 then viewt else 0 end) viewt
from temp_table
where pdt = '2023-08-17'
group by
country_id,
r_type,
c_type,
c_version;

此时结果，所有字段都是正常的！！！

此情况，在其他日期分区时，没有发生，当前运行只发生2天，但是一次的偶然意外，让我们对整体方案产生怀疑，是否可用于线上生产。。。。

想问问社区大神，这是什么问题引起的！