Figure 1. I have checked that the behavior for the north arm lock/unlocks is not due to SDB1 moving. It was not the case, SDB1 was not moving following the beam as correctly NE was misaligned.

Figure 2 shows the B1p and B4 camera with the north arm locked, and Figure 3 with the north arm unlocked. The beam is indeed more powerful when the arm is locked on B1p. On B4 it is impossible to say as the integration time setting is to low to see the beam in this configuration. There is also a ghost beam ~2mm above and to the left of the main B1p when the arm is locked, which disappears when the arm is unlocked.

Figure 4 shows the B1p and camera with the west arm locked and figure 5 with the west arm unlocked. The power on the B1p camera seems also slightly higher when the arm is locked. There is no clear ghost beam in this case.

Figure 6 and 7 show the B1p camera trend when the north or west arm is locked/unlocked. There is nothing clear visible on the power channel, but the amplitude fit is clearly going up on B1p when the arm is locked for both the west and the north arm, and for the north arm the average beam position is changing (probably due to the ghost beam).

There is probably more to learn by looking at the B1p camera images in more detail.

For future attempts increasing the integration time of the B4 and B5 camera using the GxClient may show something interesting happening to the shape of the beam (or additional spots) that could explain what is happening.

According to the simulation shown in the attched picture, the 56MHz sideband could shift because the cavity length change. 

We checked the distance between the 56MHz sideband and the TEM00 mode normalized by the FSR, for the north arm the number is 0.4528, for the west arm is 0.4416. According to this number, the simulation gives us a 3cm shorter in the west arm cavity length.