Storing the stream of tuples into a worktable to meet the intraoperator needs of an algorithm (Sort, GroupInserting, and so on), is treated as a implementation detail of the algorithm and thus is not exposed in the abstract plan.
Abstract plans expose only the worktables created for interoperator reasons, such as the self-joined materialized view. In such a case, none of the operators needs a work table. The cause is, rather, the global nature of the plan, of computing an intermediate derived table once and using it twice.
