The patch looks ok with maybe the single exception that it does a VInfo from within the fork()ed child which may (or may not) involve transporting stdout data to the MPI root process? The suggested method of using a pthread might be safer in case there are still MPI implementations out there that do not like fork(). Maybe one could add a parameter to disable parallel writing. Otherwise fine to apply.
I will apply this after the release, possibly using pthreads instead of fork.