This trivial query here doesn't seem to work:
select
count(*),
(select count(*) from dual)
from dual;
Equivalent queries work in Db2, MySQL, PostgreSQL, SQL Server, and I don't see why it shouldn't, in Oracle. The scalar subquery has no correlation to the outer query, and thus should not affect the grouping at all.
In fact, this workarounds here make it work, hinting at the above being a bug:
select
count(*),
(select count(*) from dual)
from dual
group by 1 -- alternatively group by grouping sets (())