Just as a quick word of warning, John ZuHone found this bug:
The issue in a nutshell is that fcoords is scaling between 0, 1 but not in a reliable way. I've looked into it and it seems to me that the correct solution is the one John identified: remove the scaling from yt/geometry/cartesian_fields.py .
Looking over the code, I believe most of the frontends are in place to do exactly that. I'm going to issue a Pull Request for this bug to the yt_analysis/yt-3.0 repository shortly, where I remove the scaling and offsetting from the cartesian_fields.py file. I have verified (and fixed) that RAMSES & Gadget are correctly setting fcoords and fwidth, but I was hoping to get verification from NMSU-ART and ARTIO before proceeding.
The coordinate setting inside grid_patch is correct, so I think once the octrees are verified, it's ready to go.
I'll update the YTEP to explicitly note that fcoords, fwidth must be in the domain coordinates.
PS Fixing the tests -- not sure what's up with them! -- is on my agenda for today.