J. Kissel, N. Lecoeuche

In order to reduce the number of red herrings during any given recovery, we followed the rabbit of hole of guardian warnings on the IMC_LOCK guardian which often states "IMC WFS not centered." 

This message appears, typically, after PSL incursions (rare), site power outages (rare), or computer failures (rare) when the IMC suspensions's or PSL Periscope PT alignments get lost. Because these events are rare, instituional memory loss causes confusion for folks when they see that error message and wonder if action is needed. However, once these alignments are roughly recovered (via reseting sliders on suspensions), the WFS, again typically, eventually recover their centering all on their own as the RF loops converge [there are no "DC centering" loops on IMC WFS, as there are for many of the the WFS systems]. 

Where is the threshold set and can we modify it to better reflect this knowledge?

This message is generated by a function in ISC_Library (WFS_DC_centering_servos_OK), which is called by another decorator function (gen_check_WFS_DC) in ISC_Library which creates the class (or state as we use it in guardian) called check_WFS_DC, which is called inside the IMC_LOCK guardian state "LOCKED." The decorator function, gen_check_WFS_DC, is used any where there are WFS. 
Inside the function, WFS_DC_centering_servos_OK, the threshold is hard-coded to be abs(WFS DC Signal) > 0.5 (for PIT or YAW of any WFS DC Signal) -- and note that by convention, WFS DC signals are normalized by their SUM, such that typical signals range from +/- 1.0, and centered goodness is typically considered in regions when WFS DC signals are less than +/- 0.1-0.2.

Because WFS_DC_centering_servos_OK is a function used all over the place, we can't solely change the threshold for the IMC WFS, and thus "just increasing the threshold."

So, the short answer is "no." 
However, with a bit more thinking, I bet we can catch this fish.

I attach a short (just this past time it happened during a PSL incursion), medium (including today's incursion and the past few days of lock losses), and long (all of O3) trend of the IMC WFS signals for reference.