This is an update of the plot in 76372: the high frequency (above 2 kHz) now lines up with the O4a level, but the noise between 40 and 1000 Hz is still larger now than in O4a, although a bit better than a week ago.
The first plot is done with GDS-CALIB_CLEAN: the jitter subtraction is still doing a decent job, even though it was tuned for O4a.
The second plot is done with GDS-CALIB_NOLINES, and shows the jitter peaks in all cases. There is still the same difference between now and then. So jitter is not enough to explain the lowest range we have now.