Didn't see PSU's output, but FSU's output for the same storms was correct -- I went and checked a few days ago when the issue cropped up. FSU has through the 18z runs of today's GFDL; the files needed to create the images are not made available until about 3.5hr after the actual time of the model run, with another 30 min-1 hr needed to run through all of the model products and create the images. NCEP did have some data server issues a week or so ago, while FSU had some network issues over the weekend; these temporary problems have been fixed, however.

The actual points for given storms are available much sooner from the NHC (see the Ohio State tropical sever for the text product), but the grids with the data are not. Products like GFDI and GFNI listed on that CSU model plot aren't really "new" runs of the GFDL or GFDN but are early-cycle "interpolated" runs based off of the previous (e.g. 00z interpolated models use the 18z model data with interpolation to the 00z location) model run. The late-cycle guidance, not available until some time after, is the actual true run of the model and what you see on the FSU & PSU sites.