Think of a case where all points in a binary classification problem are support vectors in an application of SVMs (not taking into consideration kernelizations of the data). How often would cases such...

Think of a case where all points in a binary classification problem are support vectors in an application of SVMs (not taking into consideration kernelizations of the data). How often would cases such as the one you identified occur in real-world data?