When escalations are enabled zabbix and something fails which causes zabbix to change the value of status item to 2 the first alert is generated properly. After escalation period the age timer for trigger starts from the beginning and yet the duration in mail is correct

.

When agent is again reachable status item stays 2 even though other items get updated which causes zabbix to send messages until infinity. The only way to get rid of the alert was disabling escalations, clearing history for status item, shutting down agent, disabling all items, enabling all items and waiting for status to turn to 2 and then start agent and wait for it to update back to 0.

Seems that there's some sort of error in updating the status of the item in database.

Also we had simple smtp check, icmp ping check and few agent checks on a host. ICMP check was working as intended with escalations, smtp check generated escalated alert but it dissappeared as status changed to 2 but zabbix kept sending escalated alerts for the smtp but not for host unreachable trigger (based on status).

When we turned off escalations everything worked as they should.