Hi Mark,
Congratulations, your data might be too clean to be processed by eddy 
Eddy’s Gaussian process expects some degree of noise in the data and if denoising works well, eddy might not – (discussion here). As far as I can tell, the error message and dependency on estimate_move_by_susceptibility are different from those reported so far though so there might be a problem with the denoised data or dwipreproc.
Might be worth going into the temporary directories created by dwipreproc -nocleanup and checking the input data to eddy and the exact commands. In particular, are there any differences in image dimensions and voxel sizes (mrinfo)? You could also try to verify if eddy works on denoised data to which you added some Gaussian noise (mrcalc denoised.mif randn $(mrmath 1.mif max -axis 3 - | mrstats - -output mean) -mult 0.5 -mult -add - | mrview - denoised.mif or similar).