For what it's worth, and it's probably very little:

Early computers had a memory that was indeed based on a trough of
mercury.  An electro-mechanical transducer generated waves at one end
of the mercurys surface.  These propagated down to the end of the
trough where they were presumably either reflected back and/or read
by another such transducer.  I suppose that any particular pattern of
waves (dunno if it was analog or digital or what) would slosh around
in a trough of mercury for quite a while.

Now, if you had a _really_ accurate timing circuit, a very accurate set
of transducers, a great deal of technical competence and a vivid
imagination, I suppose you could adapt this method to a reproducing
piano.  You'd either record the particular key that was pushed and the
force with which it was pushed (a harder push would make a wave of
greater amplitude) or maybe you could just record the vibrations of the
strings.

Mark Kinsler
Athens, Ohio USA
http://www.frognet.net/~kinsler

 [ The mercury signal delay line, and metallic salt variations, also
 [ were widely used in radar systems and cathode-ray oscillographs.
 [ It would take the acoustic wave a significant time to travel from
 [ the sending end to the receiving end (somewhat like the speed of
 [ sound in water, but sound travels faster in a solid).  If the output
 [ signal was regenerated and fed back to the sending end it could be
 [ stored indefinitely, as for computer memory.  One big problem was
 [ unwanted echoes, like trying to talk across a concrete-wall handball
 [ court!  -- Robbie