Ummm, no. A Costas loop does not decode FM stereo, but entirely other
signals.
Apparently I was not clear enough, I was referring to the post-FM
detector signal consisting of 0-15 kHz M (mono) signal (L+R), 19 kHz
pilot, the S ("stereo difference") signal (L-R) between 23 and 53 kHz
(DSB) with a suppressed carrier at 38 kHz and 60-74 kHz SCA signal
(USA) or data broadcast at 57 kHz (3x19 kHz ) in Europe.
The Costas loop would be usable for decoding this S-signal.
Moreover, by the time the pilot tone is lost the L-R differential
component is long gone.
?-)
With strong signals, the multipath propagation will act just like a
comb filter, taking out selectively some point frequencies in the
110-300 kHz broadcast signal. This might cause some severe phase shift
to the 19 kHz pilot tone, messing the stereo decoding.
Thus, it might make more sense to use a Costas loop to decode the
S-signal, instead of relying on the 19 kHz pilot, but I have no
practical experience, especially about multipath conditions.
Multipath and weak signals are quite different environments, so
different approaches are justified.
There is a similar situation with satellite broadcasts. While it is
quite easy in analog satellite transmissions to add extra audio-only
subcarriers, the reception requires that the whole 27 MHz (or
whatever) transponder can be received above threshold and _only_then_,
the audio channel can be separated.
Of course, such systems would NOT be useful for mobile audio-only
satellite reception.