I have found the work around in Cody’s post here does NOT survive a restart of the host computer. Since you said:
I’m guessing you restart and the setting in the above post are lost and same thing happens. I’ve found that if I reset the safeCall to 100, openHAB runs for weeks without running away. Typical CPU usage is in the single digits.
When it pins the CPU, follow Cody’s advise in this post above and run top in the terminal, get the pid of openHAB then run
top -H -p <pid of openHAB >
with the pid you got from top where it says pid of openHAB above
Anyhow, having to log into the openHAB console and running the commands every time you restart the host is a pain. Cody mentioned that you could probably write a rule to reset the safeCall setting on a restart automatically but I don’t know how.
I also think this is ah… kind of a kludge 
Curious @ML1982 do you run jRuby rules? What rules languages do you use?
Is this a problem with jRuby or openHAB in general? Anyone feel like explaining to me