I think in one example a Druid was narrated to be busy with a long term acrivity in camp. The perspective proposed would be that the reason they were distracted by this activity was that they didn't spot the aproaching threat. So the start of the activity the druid was busy with happened because of something that happened later. This is retrocausality.
Another way of seeing it is that you have a time period where you don't know what was happening. In order to deduce what happened you take information you have about the world that includes known facts about state
after the time period to deduce what must have happened in the time period. In this perspective I would say there are no retrocausality.
I think both perception check and the runes example can be viewed both ways. The runes example however covers a wildly longer time span. It
could be hard to argue at what point it became
qualitatively different from the perception check case. Though one cut off I have proposed might be time periods that extend further back than the last narration (last time more clumsingly formulated as the granularity of resolution).
@EzekielRaiden , do you have any example of perception checks or similar that requires retrocausality beyond the last narrated event? (Assuming we have diciplined participants that do not eagerly but prematurely start narrating before the dice hit the table)