How to get rid of accumulation of postgres temporary tables?

Question

Django

When executing recursive queries, postgres creates temporary tables that remain on disk after the query completes until the entire script completes, which is done in the form of a django management command using multiprocessing. You cannot close the connection to the database in the parent process - it is used as an iterative source of initial data.
The query is executed as a raw-query through a cursor like this:

def process_datum(datum):
   with db.connections['world'].cursor() as cursor:
        cursor.execute("SELECT ... from query_function(%s)", (datum.id,))
        rows = cursor.fetchall()
   for row in rows:
        try:
            ...
            A_Model.objects.create(...)
        except db.IntegrityError as e:
            logger.warning("%s: %s", path, e)

The `process_datum` function is called from the pool worker, the `query_function` function on the server side implements a recursive query.
Dofigischa of these requests and temporary tables litter the entire disk.
How to be?
PS
Original query: Recursive query:
select ... from features limit 1000 offset xxxx;

with recursive recursion(child_id, parent_id, node_id, path) as (
  select h.child_id, h.parent_id, h.parent_id as node_id, ARRAY[h.parent_id] 
  from hierarchy h
  where h.direct=true and h.child_id=$1
union all
  select h.child_id, h.parent_id, r.node_id, r.path || ARRAY[h.parent_id]
  from recursion r join hierarchy h on h.child_id = r.parent_id
  where h.direct=true and h.parent_id != r.node_id
)
select * from recursion

Creating request:

insert into hierarchy (parent_id, child_id, direct, path) values (%d, %d, false, %s::bigint[])

Answer 1

The problem is in the infinite recursion due to looped data.
The recursive query needs to be fixed.