BBC coincidence vs. time in fill (Run 8 pp)

During the questions time after You do not have access to view this node, I was asked if any of the h-/h+ issues could be correlated with time-in-fill because high backgrounds had been seen near the start of fills in other QA work. I commented that perhaps Jan Kapitan's investigation of h-/h+ vs. BBC coincidence rate answered this question, but I need to show that the correlation is reasonably good. I am doing that here.

Here is BBC coincidence rate [Hz] vs. day number for most of Run 8 pp during data runs which included TPC and weren't marked bad:

One can already see that BBC coincidence rates are generally indicative of time in fill. Here now is the BBC coincidence rate vs. time in fill [hours] (as measured by time of the BBC coincidence rate recording minus the time of the start of the first run of any kind during the corresponding fill):

The above plot shows that BBC coincidence rates below 300 kHz is almost entirely from data taken 2+ hours into the fills, so would not represent anything associated with backgrounds occurring early in the fills. It is important to note here that the data with rates below 300 kHz still see the h-/h+ issues.

The same information can be presented slightly differently as the time in fill which given BBC coincidence rates represent (here the error bars reflect the spread of the time distribution, not the error on the mean):

I believe this is sufficient to conclude that h-/h+ has been examined excluding runs taken early in fills via the low BBC coincidence rate cut. Feedback is welcome.