A little bit of pain is OK if the end result is good, right? Visitors to Tesly and the new Rails plugin directory might have noticed some problems accessing the site last night. Lighttpd + FastCGI + Rails decided to get a little unhappy and hosed the sites. So, I switched to Lighttpd + Mongrel + Rails and all is well.
I should have made the switch sooner, as it is rock solid. As the FastCGI listeners were spiraling to their deaths last night, I noticed periods of continued respawning. I hadn't changed anything (isn't that always the case?), so I had little clue as to the cause. Of course, by paving over FastCGI with Mongrel, I'll never know, but I can live with that. Everything is working fine now — memory use is stable, load is low and steady, and the apps are peppy.
The only problem I've had is that the exception_notification plugin's 404 handler doesn't seem to be working. Hopefully nothing's wrong with the 500 handler. :)