RHEL6 2.2.0 client sends values as active to proxy (CentOS 6, sqlite, 2.2.0) which polled (passive) by the server (CentOS 6, postgresql, zbx 2.2.0).

Client -active-> proxy <-passive- server

eth0 is a "low traffic / uptime" 10Gbit nic and works (shown as enabled, data is collected).
eth1 is a "high traffic / uptime" 10Gbit nic and does not word, GUI shows ZBX_NOTSUPPORTED

Reading both interfaces with zabbix_get works:

$ zabbix_get -k net.if.in[eth0] -s 127.0.0.1
839583684173
$ zabbix_get -k net.if.in[eth1] -s 127.0.0.1
79335063184472

So there is probably some kind of overflow in the proxy or server DB.

Network interfaces are added through LLD with the "standard" Zabbix 2.0 Linux OS template.