Indeed the glitches show up in the analog demodulated signals. So it is not an issue of the digital processing.
With an oscilloscope I looked at the I/Q monitor outputs of the LSC demodulators, including REFL11, REFL33, REFL55, POY11, AS55 while keep locking the carrier-resonant PRMI.
I saw some glitches in REFL11, REFL55 and AS55. But I didn't see any obvious glitches in REFL33 and PO11 because the SNR of those signals weren't good enough.
(some example glitches)
The attached plot below is an example shot of the actual signals when the carrier resonant PRMI was locked.
The first upper row is the spectrogram of REFL11_I, REFL55_I, REFL33_I and AS55_Q in linear-linear scale.
The second row shows the actual time series of those data in unit of counts.
The bottom row is for some DC signals, including REFLDC, ASDC and POYDC.
You can see that there are so many glitches in the actual time series of the demod signals (actually I picked up the worst time chunk).
It seems that most of the glitches in REFL11, REFL33 and AS55 coincide.
The typical time scale of the glitches was about 20 msec or so.
Note that the PRMI was locked by REFL33 and AS55 as usual.