Thanks Chris.
I hadn't spotted that the surfaces weren't identical - I thought they should have been, but as mentioned it isn't my model. Per my reply above I also found meshing errors, but for some reason it meshed happily when the cyclic symmetry was suppressed.
Your point about constraint with co-planar cut surface does make sense to me, and sounds like as good a reason as any to only use cyclic symmetry with 3 or more sections.
The 15° skew was a test case, mostly so that we could clearly see what the model was doing - we also figured that if it works at 0° and at 15° then there's a reasonable chance it will work at any angle in between!
We've got the full model running, anyway, and that's also proved used to investigate tilt (which is clearly not cyclically symmetrical). It's just become fairly slow... overnight we've run sensitivity studies on both skew and tilt angle (on two machines) and both have only reached step 6 so far!
Thanks again for the feedback and explanations.
Jonathan