Let's say you have two 20k line Python programs: those 20k lines are all different.

But the Python interpreter is identical. The Python stdlib is identical. libc is identical. The Linux kernel is identical.

Your 20k lines of Python code is a small part of this all, and most Python programs are probably >99% identical when you count it all up. Of course, this is not really all that meaningful of a number.

African elephants and Indian elephants apparently share 95% (https://archive.ph/hElph), so it seems useful. Also makes it strange that they intend to use an African elephant as surrogate mother.

Our DNA is shared more than 98 percent with chimpanzees and bonobos, our next of kin.

They are not even precise metrics. They are often made from "modified" DNA, where the parts though to be redundant are first discarded. It's more like "parts that we can compare and think are relevant match by x%".

Humans and mice are very similar to one another, so I think this is a useful metric.