I'll bet that the failure rate is closer to 2/3 than 1/2.

Question 1 at (OT) Interview questions -- your response? is based on a conversation that Ovid and I had where I described an interview question that I actually use. (Ovid translated it into pseudocode, and omitted the statement that each array can be assumed to only have unique elements.) I generally ask that question in Perl, and stand ready to explain any language construct that the other person doesn't understand. I further walk them through a 3-part question, first is to describe what it does, second is to figure out the performance, and then third is to come up with an approach (code not required) to solve the scalability problem.

About 2/3 of people I have interviewed (all purportedly experienced programmers, etc, etc) fail that test. No halfway competent programmer of my experience (tested on friends, co-workers, etc) has found it hard. (Several have been puzzled about where the challenge is supposed to be.) Knowledge of Perl does not appear to make much of a difference, even though I give the question in Perl.

And here we come to what I consider a critical point. I think that Perl experience only really matters if you want someone to hit the ground running in a very Perl-specific environment. If you use a mix of languages, then while you might like Perl, you don't necessarily want to insist on it. Looking for that is somewhat harder.

Incidentally for perspective on why companies find hiring the right person so hard, read Anne Learns To Recruit. :-)