Tricky mesh vs non-conformal grid

MN

I've got a modeling situation where I can set up (and have done so) a interface between two zones in two ways: either I can do a split volume and the like to create a single continuous connected volume set, but typically results in having some odd face meshes (but still 'ok') and one volume that can't be meshed by Hex/Cooper and only by TGrid; or I can use a non-conformal mesh and mesh everything simply with Hex/Cooper and then deal with the interface in FLUENT. Is there any gotcha's to the latter solution, given that I'm looking at only the momentum, energy, and basic non-discrete species transfer with this model? At it is, the latter solution seems to be more ideal to set up, but I'm concerned on any error/assumptions it may put into the solution.

As another question, the FLUENT manual isn't clear on this, but if I have two fluid zones separated by a physical wall, is using interface zones to connect one fluid to the wall, then the other fluid to the wall, a better solution approach than to build the model so both fluids and wall are part of one connected geometry and defining the connecting faces as 'walls' to achieve the shadow zones?