In the recently added sharding_stepdown_fcv_upgrade_downgrade_jscore_passthrough suite, setFCV can get killed by stepdowns so quickly that it can exhaust the retry limit of the mongos and it gets returned as a HostUnreachable error.

This is reasonable since setFCV may take long to run (it has to do multiple roundtrips to the shards, drain DDL coordinators...) so we should tolerate it. This is already the case in other hooks that do long running operations in the background, e.g. add/remove shards.