Figure 1. Since yesterday we have an excess in noise that limits sensitivity to 20Mpc, it has a 1/f^2.5 shape.
Figure 2. This noise is non-stationary, visible as many glitches in h(t).
Figure 3. Those glitches are short, and happening up to several times per second
Figure 4. In band passed h(t) (30Hz-200Hz) they look like ~20ms spikes. But that shape for sure depends on the choice of band pass.
Figure 5. The excess noise was already present in the two short locks that happened on Oct 14 at ~14:00 UTC and ~15:00 UTC. So if the noise is due to something breaking or changing, it is more likely to have happened on Oct 14 between 07:00 UTC and 14:00 UTC
Could it be one of the low noise switches on the input mirror or end mirrors not opening completely, and keeping the high power electronics connected to the mirror coils in a non-stationary way?