just stumbled around the net again, looking for icinga articles. this article describes a problem with mklivestatus and too less thread stack size. Previous versions used an 8MB default which was mostly bad for core forks (on each check, notification, etc). Newer versions of mklivestatus to use a smaller stack size (64KB) which would result in problems on a lot of hosts and services to be asked for, mostly crashing the core process.
But, read on over here …
http://blog.jensschanz.de/?p=1143