On Vince's suggestion, I took a closer look at the VDC data from the last cosmic run where the VDC efficiency appears to be ok. This run (#21668) was taken with cosmics on the owl shift friday morning (~00:00 - 09:30). The full run contains 169449 events. The attached plots are from the first 50,000 events of the run. At this point it looks like the efficiency is still ok.
Figure 1
Figure 2
Figure 3
Figure 4