While locking this evening, Sheila noticed that while H1 was locked in DRMI & after it went through PREP_TR_CARM, the Arm Offsets were put at wrong values (which would have prevented us from FINDING IR for later locks).  

In Guardian, we basically take an average of 1sec of data of the IN1 signals & then this avg is entered as the offset (with a "-" sign).

[From the ISC_LOCK.py script]      

• # measure ARM offsets (This is not used for the signal in PREP TR CARM?!)
• myoffsets=cdu.avg(1,['LSC-TR_X_QPD_B_SUM_IN1','LSC-TR_Y_QPD_B_SUM_IN1'])
• ezca['LSC-TR_X_QPD_B_SUM_OFFSET']=-myoffsets[0]
• ezca['LSC-TR_Y_QPD_B_SUM_OFFSET']=-myoffsets[1]

This sounds like it makes sense, but for our lockloss, Guardian took BOTH offsets to 0.0  (and the IN1 signals were clearly not 0.0).  So this looks rather dubious.  Attached is 6-seconds of data for both X&Y arms.