The problem appears on the heavy loaded systems at usage of internal checks, for example zabbix [queue].

9342:20100330:154713.734 Starting Zabbix Server. Zabbix 1.8.3 (revision 11212).
...
9350:20100330:154715.602 server #7 started [Poller. SNMP:YES]
9347:20100330:154715.634 server #4 started [Poller. SNMP:YES]
...
9369:20100330:154715.777 server #19 started [Trapper]
...
/usr/local/sbin/zabbix_server [9350]: Lock failed [Interrupted system call]
/usr/local/sbin/zabbix_server [9347]: Lock failed [Interrupted system call]
9369:20100330:155012.416 Sending list of active checks to [x.x.x.x] failed: host [xxxxxxxx] not monitored
...
9342:20100330:155137.403 One child process died (PID:9369). Exiting ...
9342:20100330:155139.407 Syncing history data...
9342:20100330:155139.439 Insufficient shared memory for ids