Hi Denis,
This looks to me an issue with sockets/DNS that give timeout? Do you have anything set under the allowed internal hosts setting?
Looking at the stacktrace, when doing a lookup of the ips it seems this list is empty (see discourse/lib/final_destination/ssrf_detector.rb at main · discourse/discourse · GitHub), triggered at discourse/lib/final_destination/http.rb at tests-passed · discourse/discourse · GitHub, hence I’m tempted to say that this could be associated to your installation (that cannot reach the ip of the sidekiq pods?)
Or check whether you are using any NetworkPolicy installed in your cluster, this can be another reason.
Cheers