Optical lever sum sees the large angle scattering from the test mass it looks at (see T1600085 for details, but note that the trans impedance and whitening gain in that document are not confirmed by measurements). This is a tiny thing compared with the power of the OL beam itself, but the scattered light is visible if you subtract the background OL beam.

I looked at the time series of OL SUM for four test masses together with MC-PWR_IN_OUT16, ASC-X_TR_B_SUM_OUT16 and ASC-Y_TR_B_SUM_OUT16 between Jul 12 23:35:44 and Jul 13 1:4:44 2016 UTC.

The first attachment is the (OL SUM -background)/TRX (or TRY), rescaled such that they are 1 at 10W. I cannot see any huge increase in large angle scatter as the IFO moves to 40W for any of the test masses.

The second attachment shows you just how the raw-ish data looks like, this is after removing the OL beam offset and taking into account the whitening gain and trans impedance. There's some difference between ITMs and ETMs but note that ITM OLs are about 5 or 6 times farther away from ITMs than ETM OLs are from ETMs.