The following test case from subselect.test has recorded incorrect result:

CREATE TABLE t1(pk INT PRIMARY KEY, a INT, INDEX idx(a));
INSERT INTO t1 VALUES (1, 10), (3, 30), (2, 20);
CREATE TABLE t2(pk INT PRIMARY KEY, a INT, b INT, INDEX idxa(a));
INSERT INTO t2 VALUES (2, 20, 700), (1, 10, 200), (4, 10, 100);
EXPLAIN
SELECT * FROM t1 WHERE EXISTS (SELECT DISTINCT a FROM t2 WHERE t1.a < t2.a ORDER BY b);
------------------------------------------------------------------------------------------------------------+

------------------------------------------------------------------------------------------------------------+

------------------------------------------------------------------------------------------------------------+

SELECT * FROM t1 WHERE EXISTS (SELECT DISTINCT a FROM t2 WHERE t1.a < t2.a ORDER BY b);
--------+

--------+

--------+

The correct result is only one row, as shown by the same example without the
indexes on column 'a':

CREATE TABLE t1(pk INT PRIMARY KEY, a INT);
INSERT INTO t1 VALUES (1, 10), (3, 30), (2, 20);
CREATE TABLE t2(pk INT PRIMARY KEY, a INT, b INT);
INSERT INTO t2 VALUES (2, 20, 700), (1, 10, 200), (4, 10, 100);
EXPLAIN
SELECT * FROM t1 WHERE EXISTS (SELECT DISTINCT a FROM t2 WHERE t1.a < t2.a ORDER BY b);
-----------------------------------------------------------------------------------------------------------+

-----------------------------------------------------------------------------------------------------------+

-----------------------------------------------------------------------------------------------------------+

SELECT * FROM t1 WHERE EXISTS (SELECT DISTINCT a FROM t2 WHERE t1.a < t2.a ORDER BY b);
--------+

--------+

--------+

In MariaDB 5.3/5.5 and MySQL 5.6 this bug is masked by a transformation that
removes DISTINCT, and GROUP BY from subqueries. However it still needs to
be investigated what is the cause of the wrong result.