This is a great article but it left out the side of this that I was really interested in which is the compute performance of these different options.

I'll take a fast deep clone for my case over one that handles all cases.