Retry on failure issue.

I think that when a serialization failure exception occurs, it should not be retried in FailoverClusterInvoker, just like a business exception.


Start by locating FailoverClusterInvoker and tracing how serialization failure exceptions are handled during retry decisions. Compare that path with business exceptions, then verify that a serialization failure stops retries while other failover behavior remains unchanged.

Written by the indexing model from the issue text.