For network discovery we have 2 pairs of conditions which we can use in different cases:
1. "Discovered/Lost".
2. "Up/Down"

We understand that "Discovered/Lost", at first glance, is more suitable to create and delete hosts than "Up/Down".
Right! But still, "Discovered/Lost" may produce some flapping - it may send undesired  notifications about host creation/deletion, which are actually false notifications, as I want to get notifications only once - when a host is really created or deleted.

To do this, I try to use different strategy - use "Up/Down" together with “Uptime/Downtime” > 7200 conditions.

This strategy is different - firstly it's more safe! Because I can be sure that temporary (up to 2 hours) not available host will not be deleted from configuration.
Ok, with such 2 conditions combined by "and", host deletion works, but ... after the host has been deleted, I get the same notification during each discovery interval, which is not suitable for me. It again misleads as I said above.
It's also true for discovered hosts, which newer have been added to configuration, which is, again, bad.

How to resolve this and get notification when host is really added deleted?
Maybe add 2 new action conditions - like "Host does not exist" and "Host does exist".
Use them together with "Up" and "Down" conditions, respectively.

So:
("Uptime > 7200" and "Host is Up" and "Host does not exist") would add host and send notification only once.

("Downtime > 7200" and "Host is Down" and "Host does exist") would delete host and send notification only once.