If a host has multiple triggers configured that are related to each other there might be cases where you get alerts for all of these triggers, resulting in a large number of alerts. A practical example would be a router with 30+ cpu's. There might be an issue that causes cpu load on one of them and you want to be alerted about that event. However, another issue might cause high cpu load on all of the cpu's and you don't want to receive alerts on all 30. Thus dependencies are impossible in this situation.