I'll admit, this is primarily focused on moving the test suite to yt-3, which no longer supports making projections for datasets with refinement factors that aren't a power of two.
I think testing the code using much more commonly used hydro parameters makes a lot of sense. That said, if others disagree, I can also fix the test failures this causes by making a patch for yt to disable generating projection tests for datasets with unusual refinement factors.
I think this is fine. The spherical infall uses RefineBy=3 because that improves symmetry during collapse (since the collapse is centered on a cell). However, in practice, nobody uses 3.
As long as @gbryan thinks that it's ok, I'm down with this. :-) Any objection?
OK, this is a very small change and no objection has been raised on the mailing list. I'm merging the PR!