I agree on several of your points. CPAN is kind of meh right now, since Python/Ruby/node.js are quickly catching up (or have passed it, even). You can quite easily find libraries existing in other languages and not in CPAN.
For me it now boils down to: sunk cost (if you're already familiar with Perl, it makes sense to keep doing it in Perl) and some features that Perl currently really excels at (variable scoping, regex integration, Unicode).
Ok, httpie is prettier and has some syntax shortcuts, but it doesn't even do a fraction of what curl or wget does. No recursive downloads, no simulating diferent HTTP protocol version, no SSL options, no even retry options! #fail
The winner will be ICANN who laughs all the way to the bank with hundreds of new gTLD registration fees (equalling tens of millions of US$). The losers will most probably be companies like uniregistry.com (or, more specifically, their investors) who will register tens of new gTLDs and later find out that they are not worth as much.
I wish I could put this comment to the top of the page. How many TLD/ccTLDs are there already? The primary interest being served here would seem to be ICANN's. As a user, I may well register some of these new domains, but I'm kind of looking at it sidewise atm.
What value does it bring? Just more dilution of the already huge namespace for the most part.
For me it now boils down to: sunk cost (if you're already familiar with Perl, it makes sense to keep doing it in Perl) and some features that Perl currently really excels at (variable scoping, regex integration, Unicode).