The spectrum of hrec during the glitch seems to contain interesting informations about the dynamic of the event.
In fig 1, one can see that hrec is almost perfectly fitted by a very simple curve:
which is the Laplace transform of:
This is not the only equation of motion compatible with that spectrum, because in the data any information about the low frequency content is missing. For example, the two time series shown in fig 2:
are both compatible with the data (fig 3). The two curves have in common the amplitude x0 of the step at t0, and the acceleration a at t0 (fig 4). The difference in the spectrum is visible only looking at the low frequencies (fig 5), but it is clear that the sensitivity of DARM is not enough to appreciate it in the data. In practice, the level of the 1/s shape at high frequency defines uniquely the amplitude of the step (and moreover put an upper limit to its duration); the frequency fN of the notch defines uniquely the inititial acceleration (just after the step) a=x0*(2*pi*fN)^2. The depth of the notch can put some constraint to the following evolution, but the arbitrariness remains high, even because the measured depth is cut by the noise floor of hrec.
Now, let's try to give a dynamical interpretation of the supposed time evolutions, in particular the one with the oscillation following the step. It seems to me that it is compatible with a fast and tiny yielding, or sliding, or adjustment, at the level of one fiber edge (anchor?). During this event the edge of the fiber goes up, while the mirror goes down (the center of mass stays still). The mirror displacement, projected along the cavity axis, is the step x0.
When the connection is restored, the fiber is a little bit uncharged, because of the displacement of its edge, so the weight of the mirror is not perfectly compensated by its elastic reaction. The residual force is the reason of the initial acceleration seen in hrec. The center of the following oscillation is neither the old zero nor the initial x0, but is shifted by the amount of the total sliding between the mirror and the fiber edge (see again fig 2).
In support of the hypothesis that a single fiber is involved in the event, there is the fact that only two lines in the comb of violin modes are excited (fig 6 - if I'm not wrong, for each fiber two modes are visible).
Lets' move to a quantitative analysis.
Unfortunately, what can be precisely measured is just the projection along z of the mirror kinematic parameters. Those are:
x0=3.75e-15 m
a=1.95e-10 m/s^2
About the sliding s0 of the fiber edge, an estimation can be done after some arbitrary assumption. In the analysis shown above, I'm assuming 1.6 Hz for the frequency of the oscillation following the event (the pitch is excited). This gives:
s0|z=2e-12 m
Considering a coupling at least 1e-2, the sliding is less than 0.2 nm (but this number could be totally wrong for a lot of reasons, first of all because all the analysis could be wrong).
In my opinion, the event could give also some interesting information about the quality of the payload mechanical response, or, from a different point of view, about the quality of h recostruction. But I leave the topic to the experts, if interested.