A team standardizes all features to zero mean and unit variance before training. A new engineer argues this guarantees the features are normally distributed after transformation. Why is this incorrect?