A laser's beam is split and then recombined, and the two path delays are adjusted so there is destructive interference at a detector. Let's also say one of the two paths has a measurable delay over the other. Classically speaking, when the laser is first turned on, the detector would be illuminated as soon as the beam traverses the short path but before the long path is traversed. Once the long path is also traversed, destructive interference occurs and the detector is no longer illuminated (and then the setup is a dielectric mirror). However, quantum mechanically speaking every photon must experience destructive interference in this configuration because there is nothing to decohere the photons. Both descriptions have problems. The classical description says that even very sparse photons somehow know how long its been since the laser's "on" switch was flipped. The quantum description allows FTL communication along the long path (one could block the long path some distance away and the detector would show an immediate change in illumination). So my question is, which is it?