Investigating the burst of triggers seen by MBTA at the end of a lock segment (see for instance the black spikes in https://vim-online.virgo-gw.eu/resources/2019-11-14/resources/20191114_mbta_R1_V1_SNRTime.png) I found a problem with the state vector.
This could be summarized in the following figure, using the example of an unlock of Nov 14, but this is not specific to this unlock, although this is one is a bit extreme.
The ITF lost lock a bit before 12:54:20 as seen by B1_DC (top right plot). But Metatron took several seconds to discover this unlock and update the META_ITF_LOCK_index only at 12:54:22 (top right plot). Hrec process the data by block of 8 seconds with 4 seconds overlap. Based on the Metatron information Hrec used the data from 12:54:14 to 12:54:22. This block of data is obviously bad and we end up with 4 seconds of bad h(t) data (from 12:54:14 to 12:54:18) that are not tag as bad by the state vector (bottom right plot). By the way, the science mode bit is still set for the second of the unlock, although that might be matter of definition (is the bit set for the next full second or only at the beginning of the second?).
It would be nice if Metatron could correctly report the state of the interferometer. It would be also good to protect Hrec against misleading META_ITF_LOCK_index and to clarify the timing of the science mode bit.
This is maybe a none problem for Detchar people. If not, it would be good to check the data quality segments, including for O3A.