We need to fix up the code throughout the system to be able to handle data faster-than-CGI processors like Plack, FastCGI or mod_perl. At this time, most of Koha can run okay, but when errors happen, they crash HARD. Some issues than need attention:
nested subroutines: mod_perl, and I think other systems, turn each script into a subroutine. If you have subroutines in subroutines (as we do in C4 and some of the scripts), and an inner subroutine modifies a variable declared outside in the outer subroutine, you will get an error "variable will not stay shared". This can sometimes (always?) be fixed by using "our" instead of "my". Or using better coding practices :)
memory leaks: some memory may not be released, so a persistent engine could bloat over time.
other issues may come up, as well.
We started a wiki page at http://wiki.koha-community.org/wiki/Plack
which documents psgi wrappers for opac and intranet along with hints how to deploy using starman
As Plack is now the recommended platform for Koha (as of December 2017) and all the bugs that this bug depend up are resolved in one way or another, can this bug itself be close?