unittest test_by_user_period() failed

Using the latest ceilometer master, with commit id:54432375ac8cfb0f58bd974ca05a44e5101bb536,

when running "tox -e py27-folsom,py27", sometimes the unittest case test_by_user_period fails with the following output:

======================================================================
FAIL: test_by_user_period (tests.storage.test_impl_mongodb.StatisticsTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/opt/stack/ceilometer/tests/storage/base.py", line 809, in test_by_user_period
    datetime.datetime(2012, 9, 25, 10, 28))
AssertionError: datetime.datetime(2012, 9, 25, 12, 28) != datetime.datetime(2012, 9, 25, 10, 28)
    'datetime.datetime(2012, 9, 25, 12, 28) != datetime.datetime(2012, 9, 25, 10, 28)' = '%s != %s' % (safe_repr(datetime.datetime(2012, 9, 25, 12, 28)), safe_repr(datetime.datetime(2012, 9, 25, 10, 28)))
    'datetime.datetime(2012, 9, 25, 12, 28) != datetime.datetime(2012, 9, 25, 10, 28)' = self._formatMessage('datetime.datetime(2012, 9, 25, 12, 28) != datetime.datetime(2012, 9, 25, 10, 28)', 'datetime.datetime(2012, 9, 25, 12, 28) != datetime.datetime(2012, 9, 25, 10, 28)')
>> raise self.failureException('datetime.datetime(2012, 9, 25, 12, 28) != datetime.datetime(2012, 9, 25, 10, 28)')

-------------------- >> begin captured logging << --------------------
ceilometer.storage.impl_mongodb: INFO: connecting to MongoDB on localhost:27017
ceilometer.storage.impl_test: DEBUG: Using MIM for test connection
--------------------- >> end captured logging << ---------------------

----------------------------------------------------------------------
Ran 630 tests in 9.134s

FAILED (failures=1)
ERROR: InvocationError: '/opt/stack/ceilometer/.tox/py27/bin/nosetests --no-path-adjustment'
____________________________________________________________________ summary ____________________________________________________________________
ERROR: py27-folsom: commands failed
ERROR: py27: commands failed

This error is not always reproducible. Sometimes it happens and sometimes it doesn't.

The problem looks like that the data returned by the query into Ming is not always in the same order as we expected in the patch https://review.openstack.org/23439.