The Hive JDBC driver used to return a result set when querying Impala. Now, instead, it gets data back but interprets the data as query logs instead of a resultSet. This causes many issues (we see complaints about beeline as well as test failures).
This appears to be a regression introduced with asynchronous operation against Hive.
Ideally, we could make both behaviors work. I have a simple patch that should fix the problem.