怀疑DynamicPartitionChannel+backup_request配合使用不再选择优先返回的下游了？

下游如果分sharding，DynamicPartitionChannel+backup_request情况下，框架没有将同shard的response去重。想问下DynamicPartitionChannel+backup_request情况下，同shard返回后框架的后续行为是什么呢？还想问下会不会出现并行的情况，目前加上backup_request会导致程序出问题。。。

Start by tracing DynamicPartitionChannel with backup_request, focusing on sharded downstream responses and whether same-shard responses are deduplicated. Reproduce the reported parallel-response behavior and determine the framework's behavior after a same-shard response; done means the behavior is explained and the reported program failure is resolved or documented.

Written by the indexing model from the issue text.