Normally the core does a snmpwalk on the monitored hosts. There are rules to overcome this problem with hugh walks and timeouts. It was discussed several times on the english aswell the german mailing list.
Check the rules related to SNMP if they fit your requirements.