In plaats van te groeperen, ga je als volgt te werk...
select
*
from account a
join (
select
account_id,
row_number() over (order by account_id, id) -
rank() over (order by account_id) as row_num from user
) first on first.account_id = a.id and first.row_num = 0