Hi Chris Rorden,
I'm preparing a large multimodal MRI dataset for shared access (period of performance started 2013). Although I have already processed the data in-house using well-established pipelines throughout the period of performance (back in the days of dcm2nii before dcm2niix), I'd like to use dcm2niix for this task since *.json files would be created/available to the curators/users of the database. 
1) Unlike the original conversion I used (dcm2nii), both the stable release and beta release of current versions of dcm2niix require an extra step for me to get the images to display correctly in FSLeyes (v0.31.2). The orientation labels are displayed correctly in FSLeyes, but the multiplanar views are not correct (although these views display correctly in MRIcroGL). I have found that running fslorient2std fixes the problem such that the multiplanar views look correct (and labeled properly). Is there a setting I need to modify in dcm2niix (i'm using default settings) to eliminate the extra step (fslorient2std)? Particularly puzzling is my observation that the 3D-T1 series displays perfectly fine in FSLeyes without running fslorient2std. However, the 3D-T2, 3D-FLAIR do not...even though they were all acquired in the same plane (sagittal).
Many Thanks for any insight you can provide.
Cheers,
Jenifer