CVR Bandwidth
I looked at specs for a couple of CVR's. One had a bandwidth of 3kHz and the other 5kHz. The microphone bandwidth would likely have been higher.
This means that a sample of the sound (think "relative" pressure) was recorded about every 100 microseconds. In that time, a sound wave travels about a couple of centimeters in air, and quite a bit more in metal. This means that the differential between microphones gives some information about the direction of arrival of an impulse, and its speed of propagation.
If the initial event was a low velocity explosion (such as a fuel tank) or an explosive decompression, there might be time for sound to get to the cockpit before the signal carrying cable was severed. If it was a high velocity explosion (C4, SEMTEX, PETN, ANFO), that sound could get there only if the explosion shock wave itself did not sever the cable, but rather just weakened the fuselage structure leading to an eventual breakup (milliseconds to seconds later).
Whether the fast rise time characteristic of a high velocity explosion shock wave could be distinguished is another question. A lot of material would be filtering out the high frequency components as the shock wave traveled from aft to the cockpit, although it might be conducted through the metal of the airframe, which would better preserve it.
Finally, the "sound" at the end of the CVR might just be the electrical transient from the cable being severed, and perhaps also from power being removed.
All of which brings up a question I've had during this very interesting thread: is there a reason we are not discussing a center fuel tank explosion? There have been a some of these in commercial jet aviation history, most notably TWA-800.