• Dividing the existing source is surely an option, but isn't very high on my priority list. There's little benefit, and a division needs care for production and installation to fit into the Emacs package. Emacs makes navigation in a big file rather easy. New features, in particular optional stuff, might go into extra files.
• Updating the docs is ... indeed interesting. There used to be a tool (written in Perl) to convert POD to texinfo, to read it directly in Emacs, but this seems to be no longer functional.
• I've already dropped all the special cases for XEmacs, Emacs below version 26, font-lock-extra, choose-color.el, and cperl-mode's homegrown fontification for multiline strings. Did I miss something? Apropos: The selection of the minimum Emacs version is a somewhat unfortunate consequence from cperl-mode being bundled with Emacs. I haven't checked yet how much effort a backport to, say, Emacs 24 would be.
• Yes, I need a test suite, and I've started writing tests for the bugs I fixed. However, I doubt that I'll write tests for the existing code unless I find bug reports. Though I would love to write a test suite in Perl (the Test:: infrastructure is lightyears ahead of ERT), ERT is mandated if you want to get stuff into Emacs. Also, it isn't that easy to query the fontification of a particular character in an Emacs buffer from Perl...

With regard to integrating with Perl, a connection to modules like PPI or even Code::ART would be interesting. To quote THeDamian: How hard can that be? But then, I'm not Damian, he's doing his sourcery in Vimscript, so don't hold your breath.