In the Zabbix v1.8.3 when there are many unreachable devices in the Zabbix system the performance of the zabbix_server decrease significantly and hence appears gaps in the data monitoring, seems that all pollers are simply waiting.
Devices are unreachable because of the several network errors:
28669:20100918:213027.762 End of get_value_snmp():NETWORK_ERROR
28669:20100918:213027.762 Item [H1:item1] error: Timeout while connecting to [H1:9003]
Maybe parameter "StartPollersUnreachable" in the zabbix_server.conf will help in this situation ?
Please consider the possibility to return this parameter back to the config file.