I've spent the day banging my head against ELF link loaders, pthread implementations and C runtime code, and the various interactions therein. While it wasn't fun, per se, it was interesting... and now that I think I've finally managed to get some indication of what the problem I'm trying to solve actually is, I'm a whole lot happier.
The actual solution (or attempt at a solution) for said problem will have to wait until tomorrow. What gets me is that, if I'm understanding things correctly, what I'm seeing is an already-known, diagnosed, and worked-around problem. Getting to that point, though, was a matter of digging through things until I realized what the appropriate search terms for my questions were...
... at which point, Google was nice enough to take me directly to the online change log for the piece of software in question. You know, the place that - if I had really been on the ball - I would have looked at first to see if there was a solution to this problem.
Bah.
Well, at least I know more about certain subjects than when I started the day...
(In my defense, I was under the impression that we were using the absolute latest revision of the software, and that looking in that direction for a possible fix wouldn't have turned up anything useful. I still should have tried, though. It's the software equivalent of making sure the power cord is actually plugged in.)