Ruby, my other favorite language besides Perl, is just as vulnerable to dependency problems as... every other language!
Ruby, Python, Perl, Java, they all have lots of packages and modules that depend on each other. So do many Linux distributions. Dependency hell is not language nor domain specific.
After all these years, one of the reasons I stick with Perl and CPAN is precisely because I can monitor and anticipate dependency problems with CPAN Testers. MetaCPAN has nice impact analysis tools too. I'm yet to find something comparable in any other language ecosystem. Hell, you can't even specify a minimum version of a module to import in Python, like in Perl's "use Module 2.3".
In that case that is most certainly not dependency hell, since no amount of dependency gyrations could fix it. The module is flat out broken on that version. Perl does however give you all the information you need to figure out on which versions of Perl the given modules work via:
http://matrix.cpantesters.org/?dist=IO-Socket-SSL-1.49
or
http://www.cpantesters.org/distro/I/IO-Socket-SSL.html#IO-So...
Armed with that knowledge you can then use Perlbrew to install that version and be on your way.
If i misunderstood, please feel free to correct me.
Name another multi-paradigm language with as broad, deep and well-tested library as CPAN.
"Unless a miracle happens, it will never return to dominance in anything other than niche domains (like quick-n-dirty commandline scripts)."
When has "dominance" (whatever that means) or popularity ever been the goal or desirable? I can't think of any language or system that is universally "dominant"; all have niches to work.
You don't get into programming because you want to do what all the cool kids are doing. Paul Graham (you might have heard of him, you're using his site) argued very well on the advantages of choosing above-average tools for above-average results. (http://www.paulgraham.com/avg.html)