Redoing the statistics.  Should I correct the 2/R "mistakes"?
   Let's not do it for now.  But I will combine H+D, P+F, S+T: 

  Tetragram frequencies around line breaks, ignoring spaces:

    cat .voyn.fsg \
      | tr -d ' /=' \
      | sed -e 's/^\(..\).*\(..\)$/\1\2/g' \
      | tr -s '\012' ':' \
      | enum-ngraphs -v n=5 \
      | egrep -v '\*' \
      | egrep '^..:..$' \
      > .voyn-nl-2-2.grm
  
    cat .voyn-nl-2-2.grm \
      | sort | uniq -c | expand \
      | compute-freqs \
      > .voyn-nl-2-2.frq

  Tetragram frequencies around blanks (spaces and line breaks): 

    cat .voyn.fsg \
      | tr -d '/=' \
      | tr -s ' \012' '__' \
      | enum-ngraphs -v n=7 \
      | egrep -v '\*' \
      | egrep '^..._...$' \
      | sed \
          -e 's/^\(...\)_\(...\)$/\1:\2/g'  \
          -e 's/_//g'  \
          -e 's/^.*\(..\):\(..\).*$/\1:\2/g'  \
      > .voyn-sp-2-2.grm

    cat .voyn-sp-2-2.grm \
      | sort | uniq -c | expand \
      | compute-freqs \
      > .voyn-sp-2-2.frq

  Comparisons:

    compare-freqs \
        .voyn-tt-2-2.frq \
        .voyn-nl-2-2.frq \
      | compute-count-ratio \
      | sort +0.0 -0.2r +4 -5nr \
      > .voyn-tt-nl-2-2.cmp

    compare-freqs \
        .voyn-sp-2-2.frq \
        .voyn-nl-2-2.frq \
      | compute-count-ratio \
      | sort +0.0 -0.2r +4 -5nr \
      > .voyn-sp-nl-2-2.cmp