I really do not know. I prefer not to make specific guesses, even lower-bound estimates, when there is insufficient information available.

I've worked in industry for 20 years. People take the same version of the interpreter, or start with existing scripts, to create new scripts. Even if there's a one rev difference, people often don't bother to upgrade. There will probably be hundreds of thousands, if not millions of Python 2.x scripts written over the next 6 years. Many will be small, of course, but it's code that will have to be checked against 3.x at some point. I threw out the conservative number because I thought it would be obvious to most people that given Python's popularity, a lot of new 2.x code will be written.

Obviously, if people were moving more quickly there wouldn't be a need to push out 2.7 support until 2020.