Figure 1, the trend in B1s2 power (and range) could be related to the CMRF degrading with time after the lock. The height of the 1111Hz (FMOD) and 3345 (SSFS) lines in DARM increases by almost a factor 2.
Figure 2, for comparison just after a lock on Aug 28 the CMRF was quite constant
To understand this more it would be really useful to a second laser frequency line (in addition to the 1111Hz) below the arm cavity pole (so somewhere between 20Hz and 50Hz) as was done during Virgo times.
Figure 3. While looking into this, I have noticed that B1s2 has a small coherence with B1 in the most sensitive region (around 100Hz). It is actually the power fluctuation seen by B1p that are not fully filtered by OMC1, so the coherence with B1p_PD1 is higher. This coherence is smaller just after the lock (blue) than 2 hours later (purple). The B1s2 spectrum increase, but the B1p spectrum remain flat (and the two are coherent at ~50%). So this means that the filtering by the two OMCs of whatever is the noise witnessed by B1p_PD1 decreases 1 hour after the lock.
Figure 4 has the same purple reference curve, but in blue is the data from end of August. At that time the filtering by the first OMC was better (B1s2 spectrum lower even though the B1p spectrum is higher), and there was no coherence.
Figure 5, The origin of the noise witnessed by B1p_DC that has a ~1/sqrt(f) frequency dependence seems to come from the 56MHz modulation. Looking at the the experiment in decreasing the 56MHz modulation depth from Wednesday, there is a clear decrease in the B1p_DC as the modulation depth is decreased with time (disregard the yellow line where some glitch in B1p must have happened). No changes in the sensitivity where seen during that experiment, as at the moment it is not a dominating noise, the coherence is below 10% for 12dBm. The explanation for why there is more coherence now than in August could be that the 56MHz lower side-band order 1 mode is close to the carrier in the replacement OMC. So that mode is not well filtered, and after the lock over ~1 hour time scale the 56MHz side-band becomes more misaligned, so the order 1 mode increases and more of it passes through OMC1.
Figure 6. It is very important to check the other side-bands (6MHz and 8MHz) by decreasing their modulation depth. These side-bands will not be seen on B1p_DC as the modulation is smaller than the 56MHz, but the filtering of the 6MHz and 8MHz is much poorer by the OMCs. The 56MHz is 20 times larger than the 6MHz and 8MHz on B1p PD1 sample data, but it is a factor 3 smaller than the 6MHz and 8MHz on B1.