rt_sleep() forever

[Malte Witt]

> As an alternative try to see what happens if your make you time hard 
> real time and let the scheduler switch it back/forth soft when Linux is 
> used.

Seems as if I was a little rashly posting on the list, sorry for that.
But your hint of trying in hard realtime mode gave me the crucial idea. 
The number of syscalls and the thread not being in hard mode when frozen 
lead me to the right position of the mistake. I will go ahead and try to 
fix it now. 

Thanks and sorry for bothering with this false-alarm.

Malte