In the sklearn.linear_models.base.center_data(...) function when normalize=True and fit_intercept=True is provided, the standard deviation of X is calculated by
X_std = np.sqrt(np.sum(X ** 2, axis=0))
I think it should rather read:
X_std = np.sqrt(np.mean(X ** 2, axis=0))
or is there any special reason why you sum here intead of taking the mean? If you just sum then while X is increased in dimension, X_std will grow also to infinity. This seems odd to me and quite unexpected.