Sometimes when a monitored host fails its items can't trigger anything, they simply go in unsupported state (think for example to simple checks like vmware ones).
In some cases when a VMWare hypervisor failed nothing was notified. So the idea was to have an item that would count the unsupported items for that host.
Now Zabbix has zabbix[items_unsupported] as internal check to count unsupported items, but it only count them on a whole system level.
I wrote a patch that I'm proposing in attachment to add a second parameter "host" to the internal check, so that it will count only the unsupported items of the host where it was applied.
I'm currently testing it on Centos7 and Zabbix 2.4.4 and it seems to work as expected.