See attachment.
We had the same situation id old releases but we weren't worried about that due to scale of the problem.
Now we have a lot of "icmpping" items in zabbix queue with no obvious reason for that.

– it is not zabbix_agentd's problem, because the item is simple
– we don't have any records in zabbix_server.log about these items/hosts from queue
– turing off / turning on is not useful, counter started from 0 minutes and item became stuck again
– non of these hosts are placed in /tmp/zabbix_server_*.pinger files

Zabbix server started at 24.04.2015 (and all items stuck at 09.06.2015).

Experimentally discovered that recreating of host is helps. Probably because of changing itemid's of all items on host.