The relationship between data distribution (i.e., topology) and data format (i.e., row vs. columnar) is confusing because it can happen in different/multiple places.

If you want replication (native or Galera), it's defined at the database level. If you want distributed or columnar, it's defined at the storage engine level. While ColumnStore and Xpand define both the format (row or columnar) and the distribution (distributed), InnoDB only defines the format (row).

To simplify things, I recommend splitting InnoDB into two storage engines. That is to say, allow it to be used two different ways (not to have two separate codebases, but two separate configurations). Ideally, "engine=InnoDB" specifies InnoDB with native replication while "engine=Galera" specifies InnoDB with Galera.

Further, this should allow for some tables to use native replication while more critical ones use Galera instead.