Sam, I am willing to suspect that government agencies with highly classified sensors won't help BEA until there is a long enough time for it to seem plausible for BEA to massage the data they have in unspecified (and never released) ways. They'd not want to release too much data about the capabilities of those sensors.
Of course, as a person who worked at a seismology lab one summer at Univ of Michigan, being able to pick an aircraft engine hitting the bottom of the ocean out of the ambient seismic noise would take REAL effort. I was involved in some field work. Explosives were being fired off Delaware in the water. I was scheduled into sites in the Ohio and West Virgina areas. Picking nice quiet sites took some effort. (Simple streams are killers if there is much water flow, for example.) One of the other team members was barely able to feel what memory declares was a depth charge like explosion of about 100# to 1000# from a site in Northern Michigan. (This was some 45 or so years ago so I'm hazy on details.) So I am somewhat skeptical about detecting bits of the plane hitting the bottom at similar distances.
(Neoflt: please note my "justification" for the 6 second figure in my previous message. The ASR33 was pretty fast for its day at 10 characters per second - 100 WPM. This is circumstantial, of course. It all depends on how the messages were stored and forwarded. Telephone lines typically ran at 75 to 300 baud, though. So we're still in the 10 to 40 cps regime.)