I looked at the HEPI tidal offloading signals from one of Friday's locks to try to extract the power circulating in the arms.

As the interferometer powed up from 2 W to 40 W, the offloading signals compensated for 4.3(3) µm of radiation-induced displacement. This amounts to 130(14) kW of power circulating in each arm for 40 W of PSL power. The uncertainty comes from (1) the difference between the X and Y offloading signals, (2) nonlinear drift in the earth tides (only the linear component has been subtracted here), and (3) an assumed 10% uncertainty in the displacement calibration.

On the other hand, if we believe that there was 95(10) kW of power in each arm given 22.5 W of PSL power, we would expect 169(18) kW of power in each arm at 40 W, assuming no power-dependent arm loss. The discrepancy between the expectation and the observation amounts to 30(20)% missing power. This is consistent with previous estimates of the recycling gain loss during power-up.

This constraint could be improved by (1) making the PSL power increase faster, so that the earth tide behavior can be better constrained, and (2) turning off the HEPI offloading and instead watching the UIM offloading.