This afternoon, we increased the pumping current of the ML from 1.774 A to 1.977 A. In the same time we were rotating the HWP at its output in order to keep the same power on the laser bench. (see plot 1). The extra power is rejected by the first polarizer of the FI in the ML box.
We did not do any direct measurment of the output power. This would have required to block the beam and it would have consequently unlocked all the PSL/INJ subsystems.  

We did that in order to see if making the ML work at a higher pumping current would improve its behavior with respect to a potential frequency noise that would cause the fast unlocks.
There is already a visible effect, the power noise has increased at high frequency and decrease at low (see plot 2).

We will let it in that state for now and study its behavior in the coming days.