Quantum Eraser Question (no Anti-Telephone BS) 
I get that the interference pattern from D0 can only be observed by taking the measurements from D1 and D2 into account. Without the this information, just combining all photons, we just see a smear.
So when we remove the entire lower part and only keep the laser, the slit, the BBO crystal and the prism behind it, the lens and D0, we would still see a smear. What happens to the entangled idler photons does not matter, let's say they just hit a wall or something.
Obviously we would still only see a smear, right?
Now I also remove the BBO crystal and the prism and of course I also have to relocate the lens and D0. This is just a conventional double slit experiment, so we should totally see an interference pattern at D0, right?
This seems strange to me. Does the BBO crystal "hide" the interference and how?
 A: Yes, the crystal hides the interference, not by preventing it, but by creating multiple versions of the interference that mask each other when added in such a way that the full pattern can be regarded as built from events that engender a concept of "which slit" the photon went through.  It does so by disrupting the phase correlations between the amplitudes of the photon passing each slit, because after it passes the slit, there is a kind of "which photon is it" choice being made in the BBO crystal, and the "which photon" can be turned into "which slit" by the appropriate measurement on the entangled partner.  That breaks the coherence, but it allows the coherence to be re-established by correlating with experiments on the "other photon" in the entangled pair in such a way as to erase the knowledge of "which slit".  
The general rule is that anything that is happening in the system that could be utilized to establish "which way" information will destroy the interference and make the "smear," so the trick is to use the entanglement to set up correlations that erase the which-way information.  So you erase what is destroying the interference.  The big "aha" is recognizing that if you see a "smear" that shows no interference, it doesn't mean there wasn't any interference in the making of that smear, it means the interferences that are there got in each other's way.  You can still tease out those interferences with appropriate correlations, because they are embedded in the full dataset, you just have to set up correlations that erase the which-way information, and you will see the different interferences that initially went into creating the "smear."  Put differently, an absence of interference actually means too many different interferences-- too much interference is no interference at all, and what we mean by an "interference pattern" can also be a "much less interference than usual" pattern.
