Hi. I found slightly different depictions of the Mandel quantum eraser experiment (1991). They are all from German sources, but I think they are pretty clear nonetheless. The first is from a German schoolbook: The second is from here: They claim that interference appears when the origins of the idler photons are made indistinguishable (second picture in the first setup, first picture in the second setup) and disappears otherwise. However it seems to me that this choice (inserting or pulling out the idler beam splitter in the first setup or the red block in the second could be arbitrarily delayed by making the optical paths of the idler photons longer, even longer than the paths of the signal photons, causing all the weird things like retrocausality or FTL communication. I know other setups (e.g. Kim et al. (2000)) don't exhibit this since they need to measure correlations between signal and idler photons, which make everything consistent with causality and SRT. Also, in Kim's setup it's not really a choice whether or not the idler photon are indistinguishable, it just happens in 50% of all cases. Are the depictions above just terrible oversimplifications, or do they really not need to measure correlations? Unfortunately I couldn't find Mandel's original publication.