Perl was in 1997 and is today a better awk+sed.

Better is to use the typesetting language

Even on MacOS, save this as `doc.tr`

     .\" Index Macro
     .de IX
     .tm \\$1^\\$2^\\n%
     ..
     .\" Content
     .sp 2
     This is the first page.
     We are talking about the history of Unix.
     .IX "Operating Systems" "Unix"
     .bp
     This is the second page.
     Here we discuss the C language.
     .IX "Languages" "C"
     .bp
     This is the third page.
     We go back to talking about Unix.
     .IX "Operating Systems" "Unix"
The index will be output on stderr with:

     % groff doc.tr > output.ps 2> raw.idx

Which will output the following that you could use bash/sed/awk/perl/... to make into another troff file.

     % sort -t '^' -k2,2 -k1,1 raw.idx                    
     Languages^C^2
     Operating Systems^Unix^1
     Operating Systems^Unix^3
(Note: `^` as a delimiter is an old Unix thing, just going really retro)

Even inserting the `.IX ...` lines on a string match in the troff file is fairly easy. But somehow they were parsing the typeset output, not sure what kind. But no need to try and track page numbers when typesetting languages do that for you.

TL;DR What I was saying is if you are using a typesetting language, use the tools in the typesetting language when possible. If you can do it in troff you can do it in LaTeX with makeindex etc...