We are monitoring several clients connected via VPN, we've made dependencies on hosts according to clients infrastructure. We have problems on recovery alerts - when Internet/VPN is down, I've made a minute delay for actions, so only router is alerted. But when connection is back, all the childrens in dependencies are first alerted with problem and few seconds later they are alerted with Resolved status. It is very annoying when you have 10-20 network devices dependent on router. I've managed to create custom recovery expression, so Zabbix is resolving after 5 minutes (5 icmp pings with state 1 (up), each checked after 60 seconds), but that doesn't work either. Furthermore, problem alerts have operational data on up.

Steps to reproduce:

1. Create copule of hosts, dependent on 1 main item

----|host2

----|host3

----|host4

1. In actions, set step duration to 1 minute, and set steps to 2-2 in operations
2. Set recovery operation for host1 for something like couple of minutes to give child items time to start replying for pings
3. Cut connection, so host1 is unreacheable for Zabbix Server
4. Get notification for host1 is unrecheable (only host 1)
5. Make connection up again, after few minutes is should resolve host1 problem
6. After host1 problem is resolved, we get problem notifications for host2, 3 and 4
7. After couple of second, we get notifications of resolved problems for host2, 3 and 4 

Result:
After bringing VPN back to live, host1 is resolved after given time, but then immidiately other host are reported with problems, but operational data in alert is Up (1st screenshot), an right away, they are reported as resolved.
Expected:
After bringing VPN back to live, host1 is resolved after given time and it should be only 1 resolved notification for this host group.

Also there could be switch or some settings for managing checked host in given order.

Maybe recovery operation delay (steps) will solve problem?