I'm behind ISP NAT. Once in a while, ISP hardware is restarted, NAT sessions are dropped, and clients are assigned new public addresses for translation. (FOPNU simply says “Reflected location: <new address>”.)

Tixati DHT graph shows this:
https://files.catbox.moe/8c1se2.png
Status says “port blocked”. Peer connections are also dropped. I stop the DHT, change global port number and DHT ID in settings, then start the DHT to get to the assumed clean state. Then everything works as it should.

It is not a big problem, the restarts happen once a month or maybe even less often, and I get to fix it sooner or later. Changing the port each hour just in case might be be a solution, but on a constantly running client it would be a disservice to other peers, and an overkill anyway. But I wonder if there is some mistake in Tixati that might make it stick to the old address, or not react as intended to the change in network conditions.