Here are the plots for itpc autocorrections vs luminosity

There is a baseline of autocorrections on the itpc due to flaky boards all the time, however at high luminosities we never see long periods of time with no autocorrects.    I suggest using the upper limits of this plot to gauge the curve of radiation induced board failures.

Of course there is not a specific luminosity where it turns on, but it looks to me like something around 85 - 90k would be the place to level the zdc counts.

I scaled the zdc RICH scalers from the zdc_mb trigger rate, so it is more accurate at high luminosity than at low.   Don't worry that the 13Khz leveling isn't at exactly the right level.