All my libraries run on Python 3 but I would recommend nobody to use them on that Python version if they can also use them on 2.x. They have better unicode support on 2.x and there are more 2.x users thus less bugs in 2.x specific code. Aside from that, the ecosystem for other libraries is bigger.
The point you raise regarding the impact userbase-size has on the library is very important in the Python2/3 split. It's similar to the divide that multi-architecture releases have to deal with, where less common platforms are more likely to have long-lived bugs just because the userbase is smaller and less likely to find and identify them.
If, as the author of an impressive collection of Python libraries, you had to make a rough estimate of when your Python 3 userbase will reach critical mass, such that they have a reasonable shot at finding those bugs and getting them fixed, how far into the future do you think we're looking?
Last time I was looking less than 5% of my users which was two months ago (up from 3% two years ago or so) used Python 3 at least going by download stats. But Python 3 support takes way more than 50% of my time :-/
