Забавный тест, вот другой, где все "несколько иначе":
~> time dash -c 'for a in `seq 100000`; do ( :; ); done'
    0m14.03s real     0m09.18s user     0m08.42s system
~> time ash -c 'for a in `seq 100000`; do ( :; ); done'
    0m45.92s real     0m33.87s user     0m15.39s system
~> time mksh -c 'for a in `seq 100000`; do ( :; ); done'
    0m18.92s real     0m12.37s user     0m10.22s system
~> time bash -c 'for a in `seq 100000`; do ( :; ); done'
    0m24.39s real     0m16.00s user     0m13.59s system

Занимательно, что твой тест в ash просто летает:
~> time dash -c 'D=1; while [ "$D" -lt 1000000 ]; do D="$((D+1))"; done'                                                        
    0m02.59s real     0m02.58s user     0m00.00s system
~> time ash -c 'D=1; while [ "$D" -lt 1000000 ]; do D="$((D+1))"; done'                                                          
    0m04.02s real     0m03.67s user     0m00.34s system
~> time mksh -c 'D=1; while [ "$D" -lt 1000000 ]; do D="$((D+1))"; done'                                                        
    0m04.80s real     0m04.79s user     0m00.00s system
~> time bash -c 'D=1; while [ "$D" -lt 1000000 ]; do D="$((D+1))"; done'                                                        
    0m09.51s real     0m09.49s user     0m00.00s system

dash-0.5.10.2
bash-5.0.17
mksh-59
busybox-1.31.1