Currently, once #syz upstreamed during moderation, all new patch candidate versions and automatic LLM replies to comments will be directly posted upstream. As the feature is still in the preview state, this is not the most optimal mode of operation.

Ideally, once it's obvious that a V+1 version of the patch will be necessary, the patch should temporarily go back to moderation, so that the person who signes-off-by the patch could iterate with the bot before it becomes public again.

Another question is what to do with the bot's comments to user replies (in case these do not trigger a new patch version, e.g. answering questions about the patch/problem). It may be easiest to just disable such comments for the upstream reporting stage, at least for now.