While I thought that the bumps observed at the end of the ringdown might be because of the cavity trying to lock itself, Jan commented that they have always existed in these measurements and their source is not known yet.
What I meant to say was that in all ringdown measurements that we observed today, the bumps were consistently part the ringdown, and that I have no explanation for the bumps. It should also be mentioned that fitting the bumpy part of the ringdown instead (we used the clean first 10us), the ringdown time is about twice as high. In either case, the ringdown time is significantly smaller than we have seen in documents about previous measurements.
We (basically I) also made one error when producing the plots. The yaxis label of the semi-logarithmic plot should be log(...), not log10(...).
I thought about why we do not find any bumps beyond the exponential fall. Could it be because we neglected fluctuations of voltage in the negative direction and plotted the absolute values?