Something that we've noticed a lot lately is that if the interferometer drops lock during the CoilDrivers state, guardian doesn't notice the lockloss for a long time. 

This was because all of the coil switching was happening in the "main" part of the state, which included many sleep commands.  Since we were switching 5 optics' coils, and each optic was about 90 seconds long, we had a solid 7+ minutes between lockloss checks.

Yesterday I re-wrote the Coil Drivers state so that the switching happens in the "run" part of the state, and I use a series of guardian timers rather than sleep commands.  So, the state still takes many minutes, but now it's constantly checking for lockloss as guardian is intended to do.

The switching has worked at least once so far with the new code, although since we didn't drop lock, the lockloss catching hasn't been tested yet.

Future work is to test whether we can do all 5 optics in parallel rather than in series, and then change the guardian to do so.