Well, I configured the “debug” level for all the cmc core sections. With the script I was able to detect the the exact time when the cpu usage of the cmc process rised to over 90%. It happens really quickly in one second from almost 0% to over 90%. Unfortunately in the cmc.log file there were no unusual messages. Also the amount of the messages did not rise. Really strange, there is no zombie process, the system and Web interface are still very responsive. However I still believe that cmc process should not use the whole core.