Machinbird, getting very deep into "noise" involves long time averages and usually something akin to a phase locked loop. With very slow Morse code your ear doesn't do as well as circuitry because your ear does not average over one second periods (or longer) very well. The Moon bounce people have some pretty clever software to play with.
Heck, even GPS is designed to work with jamming signals that are up to 100,000 times larger than the desired (P-channel) GPS signal. That's digging the signal out of noise. Again, it uses known signal properties and long term averaging to make it work.
The submarine had no chance of reception unless it's crush depth is WAY below that of anything I could find on the Internet a couple years ago. If the pinger's exact properties were known the long term average MIGHT have worked. But no two pingers will be close enough in operating parameters to characterize them to that degree of accuracy. The real weak signal people use very stable frequency references and time references to make it work.
(That's how you make something about the size of a martini olive talk to a satellite - very slowly.)
{^_^}