Indeed the contribution from PRCL has increased, and is seems to be back at the level it was at the beginning of the run.
Figure 1. Shows an offline noise subtraction where channels are considered at all frequencies the inverse of the matrix of transfer function between all the noise subtractions channels is used to be use correlated channels simultaneously (which the online Hrec subtraction is not able to do at the present). The contribution of PRCL (green) becomes larger than the contribution of MICH at 90Hz, instead of at 150Hz as it was 2 weeks ago.
Figure 2, Shows the ration of h(t) cleaned online and offline in red, the offline cleaning improves the range by about 2Mpc.
This can be compared to figure 3 and 4, which are the same as figure 1 and 2 but for Aug 5 when the range was at ~48Mpc. So better noise subtraction could bring us from ~42Mpc, to ~44Mpc, but there is still 3-4Mpc missing.
Figure 5, compares the offline cleaing between Aug 11(red) and Aug 5 (black). There is clear higher shot noise now, which should be due to poorer optical gain, but this is not explaining the remaining difference in the 80Hz-200Hz bucket.
In any case, switching back the cross-over between SPRB_B4_56Mhz_Q (MICH) and SIB2_B2_8MHz_I (PRCL) subtraction in h(t) from 150Hz back to 90Hz where it was since the beginning of the run, should improve the range by ~1Mpc.