While checking the reprocessed data of the first month of O4b, I saw an event on 2024-05-04 05h03 UTC, where h(t) has larger values over several minutes.
The same behaviour occured in the online data, so the problem does not come from the reprocessing.
Plot1 shows the reprocessed h(t) trend data (upper row) and the online h(t) trend data (lower row) around this event.
Plot2 shows the spectrograms of DARM, hoft_raw and hoft, where we can see a wide band strong glitch present in hoft_raw but not present in DARM. Also, we can see in hoft a larger noise over several minutes, that is due to the noise subtraction whose TF was biased by the presence of the strong glitch.
Plot3 shows that the origin of the glitch in hoft_raw is one wrong sample in the signal Sc_NE_MIR_Z_CORR.

This event occured while the ITF was in science mode and the state vector saw it
(its value was not 4095 for one second at 05:03:24 UTC but hoft_raw was wrong already 8s before that time because Hrec uses FFTs of 8 sec).

I found no trace of this event in the logbook, in the DMS archive or in the logfile of Hrec.
So, maybe some additional monitoring (in DMS and in Hrec) is needed for such an event (wrong sample in correction signal).
Moreover, even if such event seems rare, some protection may be needed in Hrec
- to check the control signals before using them
- to prevent bad data over several minutes by protecting the noise subtraction process from strong glitches present in hoft_raw or present in the noise channels used.