Hi,
in the meantime the 5.0 longitudinal stream has been fixed and tested.
See: http://surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalChangeLog for links to download the fixed binaries and updated recon-all script.
Note, that we recommend to wait with any large scale longitudinal processing for the 5.1 release this fall, that will include several improvements (see link above).
Detailed bug description: mri_robust_template and mri_robust_register make use of a shared FreeSurfer library to transform/resample images. A non-related, last minute, 'bug-fix' in that transformation library introduced a new bug: Instead of taking the target geometry information from the LTA, it was incorrectly copied from the source image if no specific target is passed. This broke the robust tools and thus the longitudinal stream.
Cheers, Martin
Enjoy the free longitudinal surfin' ...