One of the potential culprit for the degradation of the green signals that has been observed these last days could be the Stimulated Brillouin Scattering into the 3 km long fiber. The theorical threshold is about 8 mW. When we installed the fibers the power at the input of the fibers was about 7 mW (entry https://logbook.virgo-gw.eu/virgo/?r=49235) However, reading again the log entries, it seems that at the end of the day the power injected was actually higher ( https://logbook.virgo-gw.eu/virgo/?r=49425)
In theory the power picked off from the LB towards the other subsystems is kept constant and we readjsut it whenener we increase the power of the fiber amplifier. After the slight increase (10%) of Tuesday ( https://logbook.virgo-gw.eu/virgo/?r=53212 ), we did not not adjust the power of the pick off. Moreover after a year of functionning and the different power adjustements on the PSL side, this power may have varied.
This morning we went into the DAQ room to measure the power at the input of the 3 km long telecom fiber. We measured about 12 mW for each fiber.
We added a 90/10 coupler at the input of the splitter box in order to reduce this power. From 26 mW at the input of the splitter box, we went down to 16 mW (extra losses not investigated) and we finally got about 7 mW at the input of the 3 km fibers. 
The 2 W amplifier at north end did not restart by its own, we went there to restart it.
Apart from that both sources restarted without any problem. We will monitor their good functionning during the next days, they may suffer from the power/polarization changes of the seed.
We will also monitor if it has an effect on the glitches.