Steps to reproduce:

1. You have two devices connected, a router and a switch (Zabbix -> router -> switch). If the router is unreachable then the switch will be unreachable too, there is no need to notify about both of them, just the router (like in this example
2. Create an Event correlation rule to automatically closed the ICMP unreachable problem of the switch when the router becames unreachable.
3. Fire the ICMP unreachable trigger of the router to test event correlation rule.
4. Go to Problem -> Show Recent Problems and check the switch ICMP unreachable problem is closed.

Result:
When a new value of the item ICMP ping of the switch is received by the Zabbix server, the ICMP unreachable problem is triggered again and inmediatly closed (again) by the event correlation rule. And this repeats over and over until the router problem (parent) is resolved.
An unnecessary load could be put on the database if a problem is kept open for a long time or with a event correlation rule applying to a lot a devices.

Expected:
When a problem is closed by an event correlation rule, that problem should not be created again (when a new value is received and evaluated by the trigger) until the parent problem is resolved. It should be withhold the same way notifications of dependencies are in the case of trigger dependencies.