I had an issue where I had 200+ SNMP items on a single host and for some unknown reason Zabbix 1.8.11 was unable to successfully monitor all the items. I kept on getting errors like "SNMP item [ifOutOctets13] on host [Cisco 877 ADSL Modem] failed: first network error, wait for 15 seconds", which of course resulted in the other items on that very same host being queued for longer than 10 minutes, and thus leaving massive gaps in my graphs.
The thing that bothered me was that snmpwalk NEVER had any issues querying my SNMP items, yet Zabbix failed to even go successfully through all the items 1 time.
I really tried everything that you can tweak for Zabbix and SNMP: using proxies, changing the StartPollers and Timeout in the config, etc.
Since I was monitoring so many items, I never bothered to carefully look at the zabbix_server.log file, which would've pointed out to me that it was only 3 items out of 200+, which had wrongly configured SNMP community or SNMP ports!
So once I've fixed my 3 SNMP items' community and port, these errors stopped and my Zabbix queue cleared up!
So my advice would be to double check those individual items, by running snmpget (part of yum's net-snmp-utils package) from the monitoring machine with the same parameters that you configured for the item. Make sure of the community and port too!
If that didn't help, then I would suggest that you start debugging Zabbix's load, by importing the Zabbix Server template that can be found at: http://blog.zabbix.com/monitoring-ho...esses-are/457/
I hope this is helpful!