According to Fig.3 in the original post (klog#33198), optical gain is increased as 1.5dB (246.601dB -> 248.094dB). So current sensitivity seems to be underestimated as 1.5dB not overestimated.
The original post is a talk in my sleep.

Changes from Feb are as follows.
IM_LOCK_L (Fig.1)
- notch filter was switched from FM2 to FM9
- notch design in FM10 was modified

MN_LOCK_L (Fig.2)
- notch filter was switched from FM8 to FM9

LSC_OMC_DC (Fig.3)
- FM2 (+2dB) and FM3 (-2.08dB) were added.

So there is no change in the overall gain due to the digital filters and 1.5dB difference should come from an increase of optical gain or actuator efficiency. Temperature of IM was ~10K different between Feb (T1 cursor) and yesterday (T2 cursor) as shown in Fig.4. cross hair shows a time around previous best(?) sensitivity discussed in klog#33015 and klog#33099. Though I have no idea to the temperature dependency of the actuator efficiency around 80K, we may be able to predict that a 1.5dB change comes from the optical gain or actuator efficiency by the comment from CRY folks. (According to Komori-kun's comment in the morning briefing, mismatch in the shot noise level is consistent with an assumption that 1.5dB change comes from one in the optical gain...?) If it's difficult to conclude by a discussion only with current results, it had better to do the full calibration after IM temperature becomes stable.

BTW, the full calibration process requires non-short time (~2-3 hrs. per once). So frequent execution of the full calib. process steals a lot of time from commissioning activity. From the view point of the time sharing, improving the stability of mirror temperature and cry-cooler becomes important if we need to take care about temperature dependency of the actuator efficiency around current temperature.