Related to (but different from) #6136: float64.__str__ can truncate different floats to the same string, even when printoptions is set to a high enough precision:

In [12]: np.set_printoptions(precision=1000)

In [13]: x = np.float64(0.20000000000000004)

In [14]: x > np.float64(.2)
Out[14]: True

In [15]: str(x)
Out[15]: '0.2' # oops

Note that repr(x) correctly gives '0.20000000000000004'.