Dariusz Jędrzejczyk

This works as designed. We have a separate configuration for domains and it's undesired to have a cluster with both IPs and domain as the endpoints.

`Scheduler` restarts make it harder to implement having `onDispose()`. We should consider imposing a contract there to disallow restarts.

The deprecation of restarts is already happening via deprecation of the `start()` method and introduction of `init()` via https://github.com/reactor/reactor-core/issues/3202

@haraldpusch would you consider contributing a PR for this as it seems you have an idea where the fix can go?

To my understanding, the current behaviour violates rule 6 of RS spec for Publisher: https://github.com/reactive-streams/reactive-streams-jvm#1-publisher-code as it handles a cancellation signal after a terminal signal. This is in contrast with...

Thank you for the report. We are aware of the issue and it might require some significant design decisions to fix. Unfortunately, for the time being, I have no immediate...

Thanks for the suggestions @plcarmel. We had a bit of a discussion around this problem and the background of the current state of affairs. It looks like we can try...

@plcarmel please have a look at the latest 3.4.x snapshot and test with your use case. Feedback is welcome. The implementation resides in `FluxPublish` instead of `FluxReplay` as the latter...

@plcarmel would you like to submit a PR with updates to the javadoc so it reflects what you'd expect to read? I guess I wouldn't do a better job at...

That sounds like a good idea. It still does not honour the TTL as other cases of non-zero history, which would reset themselves after the TTL. But it is an...