In this real life example, the bigger brace has a lot more freeboard* than the smaller one does. groff -ms <<! .nr PS 12 .nr VS 14p .LP .EQ left "" rpile {sum from j=0 to j sub 0 v(j,2i) above sum from j=0 to j sub 0 v(j,2i+1)}~ right } = left { pile { 1 sub "" , above 0, }~~~ lpile { i=i sub 0 above i sub 0 <i< i sub 1 } .EN ! I infer that the brace heights are rounded up to multiples of 1v and wonder whether the parts couldn't be slid up and down less jerkily to provide a consistent amount of freeboard.
* Borrowed from nautical usage, freeboard is the amount of brace that extends vertically beyond the embraced text. Doug