We don't test against Py 3.3. Like lmfit 0.7.4, it is pretty old.Python 3.4 was released in early 2014. My view is that Python 3.0 to through 3.4 are not interesting or useful at this point. I would have no objection to dropping support for Python 3.4, and cannot be very sympathetic to people who claim to be stuck with Python 3.4 or earlier. FWIW, our policy is also that scipy should be newer than ~2 years old. We don't have the manpower to support legacy systems. Not upgrading for 4 years is not a good strategy.
That said, I am slightly sympathetic to people stuck with Python 2.7, but lmfit 1.0 will almost certainly require Python 3.5 (or perhaps even 3.6) or higher.
Python 3.7 is pretty new. The latest release of lmfit (0.9.11) was released within a week of Python 3.7, so there was not much testing done with it at the time of the last release. But we do test against Python 3.7 and I amn not aware of any problems.
--Matt