At 2:18:50 UTC the IFO reached LSC_FF and I engaged the Intent Bit as a test, since Jaime had just had me load new code into the ISC_LOCK guardian, that will unset the Intent Bit at lock loss.
NOTE: This lock stretch IS NOT at about 60Mpc - the calibration is KNOWN TO BE INACCURATE, and is being worked on.
ETMY ESD is saturating constantly, so while I set the Intent Bit as a test, this lock stretch is of QUESTIONABLE USE for data analysis.
2:32 UTC, lock loss and Intent Bit is still green, calling Jaime.
Hopefully addressed. See comment to previous post.