The GFS and even some members of the GEFS tend to not like pattern changes that are establishing in the 8-16 day window.  You’ll generally see at least 1 if not 2 of the Op runs and about 30% of the ensembles on a given day of model cycles try to break it back down to whatever the regime you’re coming out of is.  That’s been evident on the GFS the last 72 hours post 280-300 hours on some runs.  The ensemble average though is still generally showing the overall switch to the +PNA establishing