corrcoef[1, 0] != corrcoef[0, 1]

On Python 3.5 (from Anaconda) with numpy 1.10.1, the following test fails:

```
import numpy as np
import numpy.testing as npt

np.random.seed(100)  # Only happens sometimes! 
t = np.arange(1024)
x = np.sin(10 * t) + np.random.rand(t.shape[-1])
y = np.sin(10 * t) + np.random.rand(t.shape[-1])
cc = np.corrcoef(x, y)
npt.assert_equal(cc[0, 1], cc[1, 0])
```

Is this by design? 
