Simon Baslé

hi @Tariod ! Thanks for the PR (I knew it would come one day 😉). This one is quite huge and won't probably make sense until 3.5+, so I'm turning...

@Tariod no problem, thanks for the update. I see that two tests that are likely related now fail on the latest commit (publishOn tests), so it will need some investigation....

I was running all Junit tests under reactor-core.test module for coverage (including loops, that are usually executed during releases), and I see some issues: - `FluxPrefetch` only has 40% coverage...

> * `FluxPublishOnLoop` running for 11 minutes ! I'm running the `main` version to compare, this might not have anything to do with your change @Tariod 😉 edit: it runs...

cc @marcingrzejszczak

deferred to M3. we need to reach a broader consensus on the utility of this feature, from downstream stakeholder projects

yeah this is something that we are still monitoring. we cannot yet answer many of these questions. plus, a good portion of these seem a bit skewed towards a pretty...

I think it can be done by capturing the requests made BEFORE the consumer is set in a `EARLY_REQUESTED` atomic, detecting the consumer is _being_ set by CASing `-1` into...

IIRC this is a blanket behavior of window operators, that have to deal with currently opened windows on top of the main subscriber when propagating an upstream error. Buffering operators...

It would be a breaking change, but one possible solution to differentiate between errors in the window consumption (ie. part of the `flatMap` function) vs errors propagated to an open...