Sorry for the slow reply. From the VELO experience, I think the main thing I would suggest to consider (if you haven't already) is if you want to apply any requirements on the PV positions used. And probably good to test it on data with SMOG.
Hi Christina, thanks for the investigations.
It is the same event being injected over and over again. The reason why in some events different numbers of tracks/vertices are reconstructed is under investigation, but is likely due to the fact that the event which we are injecting is quite busy and is pushing the limits somewhere (ASIC or the firmware). Anyway we have seen that the variation of total clusters between events is ~10 out of ~1700, so I think it is rather unlikely this has such a large effect on the rate.
Anyway, in case it might be helpful we can also try with a less busy or different event.
Nathan Philip Jurik (1fe1ef3b) at 04 May 15:22
scripts for closing monitor task
Nathan Philip Jurik (e027b51c) at 20 Apr 17:50
For reincluding the closing monitors with rhel9
Nathan Philip Jurik (f0b12d9f) at 20 Apr 17:39
For reincluding the closing monitors with rhel9
... and 6 more commits