I've set up 2 hosts, that monitor the same SNMP host, one directly form server, one from proxy.
The unavailable/unreachable settings are default:

# UnreachablePeriod=45
# UnavailableDelay=60
# UnreachableDelay=15

Each host has 600 items with update interval 30 minutes.
If the snmp daemon is shut down, this is written in log:

23083:20190607:163217.557 SNMP agent item "snmp[snmp1]" on host "snmp-test-server" failed: first network error, wait for 15 seconds
 23084:20190607:170217.810 temporarily disabling SNMP agent checks on host "snmp-test-server": host unavailable

Notice, how host becomes unavailable exactly 30 minutes after the first network error. While the host is in this unknown state, other items are also not checked at all, but the host is shown as available.
The same thing happens if Zabbix agent is used. In this example, the update interval is set to 5 minutes for all items on the host, but 1h for the discovery rule:

  1488:20190607:165950.398 Zabbix agent item "net.if.discovery" on host "test" failed: first network error, wait for 15 seconds
  1490:20190607:170452.030 temporarily disabling Zabbix agent checks on host "test": host unavailable

Another strange thing is, that the queue grows for the proxy, but not for the server. The host is monitored from the proxy and server, when it goes down, only the proxy reports delayed data: