BUG: Large MultiIndex-ed series fails on slicing

[max-sixty]

With a large MultiIndex-ed series, slicing doesn't fully slice on a 2nd-level.

I've been struggling with this for a while, any ideas greatly appreciated...

In [135]: ints = (pd.np.random.rand(150000)*1e7).round().astype('int')

In [136]: index=pd.MultiIndex.from_arrays([list('a'*50000 + 'b'*50000+'c'*50000), ints])

In [137]: series=pd.Series(np.random.rand(150000), index=index).sort_index()

In [138]: series
Out[138]: 
a  66         0.763398
   171        0.886328
...
   9999825    0.329101
   9999933    0.463330
dtype: float64

In [139]: series.loc[(slice(None), slice(5000000))]
Out[139]: 
a  66         0.763398
   171        0.886328
   186        0.315654
...
   9720405    0.712525
   9925259    0.016771
   9978827    0.555781
dtype: float64

... the slice clearly has values both above and below 5,000,000. It does do some slicing though, just not fully.

Or am I making a rookie mistake?

pandas: 0.18.1

[jreback]

dupe of #12896

there is a bug in ten slicing logic