Chromatic calls us to Remove the Little Pessimizations, to sand off the tiny, rough corners of our software that individually are small, ignorable distractions. I've read somewhere that each barrier to entry reduces the number of users by some percent - and it is a good model to think about it. Each one reduces the number of possible uses and users and the effect is cumulative - hundreds of small barriers make the tool useful in only a narrow niche, but even just one small barrier can make a tool not usable in a particular setting. The question is when fixing these little pessimizations becomes needlesly pedantic.
I believe that fixing CPAN installation issues, each one usually a small issue occurring in some special cases is needed (at least for the most popular modules) if we don't want to confine CPAN users, and since CPAN is the killer application of Perl, then it follows that also Perl users, to a narrow niche.
Another thing is how off-putting are the pessimizations that seem gratuitous, you immediately feel neglected and you imagine how the arrogant developers just don't care about your case. Then there are also pessimizations that look small if you have a big project - but are impenetrable barriers for small projects - this is how Perl lost the place of the most used language for WWW programming to PHP.