I don't think here is an issue.

See also ZBX-8475

ZBX-8528 could explain some details of this issue.
When Andrey has created this issue I recall we discussed zabbix server behavior in unstable network where many network errors logged for snmp checks.

I still don't see anything wrong in first sentence of issue description.
Remember also about item update interval and other snmp parameters.

To be a bit more precise, function __config_snmp_item_compare() in dbconfig.c is used for comparing SNMP items in the queue. Since for SNMP bulk purposes all relevant parameters have to be the same, we can only take items from the queue which are the same according to this function.

Andrei, taking the above into account, could you please elaborate more on the problem and how to reproduce it?

If you have an environment where this issue is reliably reproducible, could you please show a tcpdump of the traffic between Zabbix server and one particular device that shows that it does not work according to https://www.zabbix.com/documentation/2.2/manual/config/items/itemtypes/snmp#internal_workings_of_bulk_processing ? Please capture traffic right from Zabbix server restart and attach the *.pcap file here.

For instance, a command line like the following could work:

# tcpdump -npU -w traffic.pcap 'host 192.168.1.1 and udp'

Also, what kind of devices are you monitoring and how many SNMP items do they have?

It seems it is dupe of ZBX-8528