Any time that there is either a big measurement, or any kind of glitch (including when the squeezer beam diverter is opened / closed, or our regular mysterious glitches), the ADS dither lines around 20 Hz get drowned out, and the ADS starts to pull the PRM and the arm SOFT degrees of freedom to weird places. 

We've put turning off the ADS loops into the NLN_CAL_MEAS state, which we (should be) going to for any measurements, but if there's a glitch or other unexpected noise that lasts for more than a few seconds it's possible that we won't notice (to turn the ADS off) and we'll lose lock.  In the attached screenshot of a glitch time versus a quiet minute later, we see that the lower frequencies that we'd been trying to go to aren't as effected by the glitch, so we could perhaps survive some fraction more locklosses. 

Having said this, let me emphasize that most glitches are brief enough that the arms will get a bit of a wobble, but then we'll have good ADS error signals again, and be back in business without any trouble or lockloss.  Moving the dither lines should't become our new #1 priority, since lockloss  from them happens fairly rarely.  However, it should remain on our to-do list for during the run.