Do passive locator systems such as RECCO (for finding skiers caught in avalanches) function underwater?
No, and they sometimes don't function under several feet of snow, depending on the attitude of the slope, wetness of the snow, so forth.
Any beacon signal is a last-ditch effort - broken beacon, unfavorable conditions.
Seriously -
Having a transponder detach from a frame is only good in a military situation where you can assume several other craft will be in the same general area and can actively locate the transponder.
Trying to transmit all the parameters recorded by the recorder(s) is a nice idea, but unrealistic at this point mainly due to cost. Unrealistic also considering bandwidth - not that a "perfect" connection would not be able to transmit all the data, but assume that an imperfect connection would not be able to transmit enough of it faithfully.
What could be done is this - very short bursts of data which include indicated position and perhaps GPS position, transmitted every 2 - 5 minutes.
Just having that data would reduce the search area tremendously. If you used 2 minutes as the design "center" value, then you'd be able to tell within a few dozen miles exactly where a flight stopped relaying data, and from that you could narrow the search area greatly, perhaps to less than 100 miles.
The "guess" zone for 447 was huge initially, and was eventually refined to another "guess" based on ocean currents, but the true location hasn't been pinpointed yet.
If one could have started the search assuming 100 mile accuracy, the boxes might be in the lab by now.
Point being, it's not so much
what you transmit, it's
where and
when you transmit.
ACARS hasn't been too much help. The pingers on the outside of the boxes haven't been too much help.
What would have really helped would have been a simple lat+long from the aircraft, updated for as long as possible, even transmitted through ACARS.