One of my brothers (I have four, two dead, two living, and I never knew about 'em growing up) is a project manager in London and he oversees a bunch of Java programmers in a grid computing project. Curiously, while he's not overly fond of Java, he told me quite bluntly that he won't approve Perl in his shop.
The problem is that he knows he could get a crack Perl programmer to come in and do wonderful things, get it done faster than Java and then leave. That's where he's stuck. He claims that finding competent Perl programmers is a tough thing to do. With Java, he feels he can pull just about anyone off the street and even if they're only average ability, he can contain the problem. Not so with Perl.
There's also the perceived problem that many of the large scale issues he works on don't have Perl software packages to support them. Grid computing, robust ORMs (many of those for Perl have all sorts of subtle issues in large projects with multiple concurrent users), threading issues, and so on.
I think there are several problems here, all of which are valid. Most programmers (any language) are awful and since there are fewer Perl programmers, we wind up with fewer good Perl programmers. Perl's security model, outside of taint checking, has never been well-defined. Further, due to the extraordinary richness of Perl and how easy it is to specialize in a particular "strength" of Perl, a good programmer in one area can be dangerously incompetent in another. This can make it very tough to evaluate a potential candidate.
Regardless of whether or not you agree with this perception, it's the perception that many folks share. If the perception is right, we're shooting ourselves in the foot by not working on the underlying issues. If the perception is wrong, we're shooting ourselves in the foot by not correcting the perception. Of course, how to implement those corrections is the tough part. Many don't care (this is the "I have a job, screw you" attitude), so they don't help. Others don't agree on whether or not the perception is correct, so they don't help. Those who agree on the nature of the perception don't agree on the solution, so they don't help. Those who are in full agreement often don't have time, so they don't help.
Meanwhile, another Java programmer just got hired.
Update: I think it's worth pointing out that many of the folks here on use.perl are good enough that I suspect they've forgotten just how phenomenally bad/dangerous incompetent Perl programmers can be. We don't give 'em a rope to hang themselves. We give them a Gatling Gun with a hair trigger and a target rich environment.