Any reason to use multiple non-nested tf.GradientTape() calls?

Question

I have looked at the documentation of tf.GradientTape() and understand that calls can be nested to track higher-order derivatives.

But is there any reason to