There was definitely enough space on the device where the locks are stored (default /usr/local/apache2/logs/). I tried to explicetely different Lockfiles using the LockFile-directive but this did not help. I also tried a non-default AcceptMutex (flock) which then solved the acceptlock-issue and ended in the rewrite_log_lock-issue.
Only reboot of the system helped out of my crisis.
Solution: There were myriads of semaphore-arrays left, owned by my apache-user.
ipcs -s | grep apache
Removing this semaphores immediately solved the problem.
ipcs -s | grep apache | perl -e 'while (
