I have fixed this. The blog is now running entirely through PHP's "FastCGI
Process Manager" combined with nginx.

This should make it more stable. If it crashes again, we should file more bugs
(or perhaps just stop hosting the blog ourselves).