Having spent a few years in the ASW community (though it's some years ago), I would like to comment on sonar search optimization in the military sense, and the impression Hollywood and Tom Clancy have given of magical sound detection capability in military submarines using sensors optimized for military use.
The whole point of tuning your gear, and positioning it in the water column when you are using a towed array, is to
1) put it in a good direct path posture for some missions
2) take advantage of various sound channels, sound layers, or convergence zones in other missions.
This sensor positioning is predicated on searching for "things" between the surface of the ocean and about 3000 feet, things which try to hide in the "shadow zones" between various water layers in a given area. I use 3000 feet since it is well beyond crush depth for every submarine I can think of in all navies, and roughly where the main thermocline is over come by the isothermal layer.
Thanks to physics, this depth range represents the zone where sound velocity undergoes a series of changes for a variety of reasons. Deeper than that, the velocity gradient takes on a predictable profile.
http://usna.edu/Users/physics/ejtuchol/Chapter5.pdf
See page 5-10 for a nice picture of the whole ball of wax for very deep oecan.
As the ocean gets deeper, the influence on sound propagation with basically constant temperature at some point around 3000 feet is water density. The other factors (like salinity) tend to have reached a constant by then.
This "roughly homogenous" medium (water, cold, deep) removes the various exotic sound channel features like the deep sound channel and the convergence zone, which take advantage of the changes in sound velocity from shallower water to deeper water.
With the pinger at depths in excess of 5000 feet, the long range signal reception enjoyed against manned submerged "things" most likely cannot be taken advantage of. (Caveat .... depending on what the CZ and sound channels are doing in that locale on that day ... it can vary a bit).
With the source well below the main thermocline, my back of the napkin estimates are that you are stuck with a direct path propagation ... IF you can get your sensor below some of the last sound channels at depth (and thus a possible difractive layer). If the sub is trailing a towed array at, say, 500 or 700 feet, he's not going to be able to trail it at 3000 feet ...
Granted, it's been a few years since I had to chart sound velocity profiles from at BT trace. That said, I cannot recall any military tactical hydrophone optimized for detecting "things" below 2000 feet, even though sounds of many sorts would follow varying paths from various depths to reach the hydrophones.
I am pretty sure that SOSUS (if it's still in operation) would not be listening in the proper layer of water to hear the pinger. I think the pinger's sonic waves were likely "bent away" from a hydrophone array at tactical depth. The best one could hope for is the pinger being in a zone of "half channel," but I don't recall if there is a depth limitation to that phenomenon. I seem to remember the limit being less than 10,000 feet, and half channel being somewhat available in the Med during some seasons ... but it's been a few years.
For those more current in acoustic sensing, I'll accept corrections to my recollection on how all of that sonar detection fits together.
It amazed me to learn that post mission acoustic analysis of the tapes from the submarine showed that a pinger signal had been detected. I suspect that "which direction was it from us at the time?" was very difficult to confirm. Had they verified a detection 'in situ' during the mission, the captain might have been able to use some simple tactical maneuvring to arrive at an AOP and expedite the search and location while the pinging was in progress.
EDIT: the more I think of it, the less likely it is for half channel to have been in operation, as I think that requires a surface layer temp roughly equal to bottom layer temp, little to no mixing layer and a few other things that memory does not serve up.
This crash was in June, so half channel not likely in the tropics.