I haven't used FiPy but only OpenFOAM, but I think some points on both homepages suggest it strongly depends on what you want to simulate.
OpenFOAM already has some solvers imlemented (mostly CFD, since your profile states computational plasma physics the mhdFoam solver may be what you want or at least serve as basis, here's its user group), but can be expanded to solve other FVM-feasible PDEs, while in FiPy you'll probably have to implement the equations from scratch anyway - and in the latter case, yes, Python is a lot more comfortable than C++, at least to me.
There are some Python codes that make life with OpenFOAM easier: PyFoam to help setting up cases, and pythonflu, a wrapper to write solvers in Python (untested, I don't know how that performs).
From my experience, setting up a case in OpenFOAM is rather tedious and you will most likely end up writing scripts to automate the mesh setup etc anyway, so I don't know whether FiPy would actually require that much more effort even if you used one of OpenFOAM's already implemented solvers. And should you choose to use Python for that scripting, I suspect you can write most of the code in a way that the decision on FiPy vs OpenFOAM boils down to a difference of less than 100 lines of code, i.e. it shouldn't be too difficult to change your mind later on, especially if planned beforehand.
The more important points are of course, how much time does the first time setup take and how well does it perform. I couldn't find any comparisons on either though, so you'd have to compare them yourself...
Personally, had I not already used OpenFOAM before and I would need to simulate a system for which OpenFOAM does not already provide a solver, I would most likely try FiPy first due to my preference of Python over C++. But as mentioned, together with PyFoam one could probably come up with a rather generic framework to compare the two... If a solver of OpenFOAM could be used though, I'd probably use that one.
