If I want to get a BatchUpdateException then is it typical to have to catch the RollbackException get it's cause and then follow it's sqlexception to get the BatchUpdateException so you can then look at the update counts and then in my case kick out the offending record from the batch (a collection) and then try again?

Usually, when you get an exception, you have to close the Session and notify the caller of what went wrong.
Retrying makes sense for some specific use cases like optimistic locking or a timeout.
When the incoming changes violate a constraint, retrying does not make sense since that request will always fail.