Below is a first look at the minbias trigger id 580001 with fast offline.  Looks pretty good for |vz|<30 cm (distribution of refmult^{1/4} should be roughly flat, and it is),
but only 85% of the events have a vertex and only 42% have a |vz|< 30 cm.  This is because the zdc, which is more sensitive to peripheral data, has a wider vertex cut. I also looked at abs(vz) for those events that have a vertex, vs. refmult, in order to get a feeling for how wide the vertex is in peripheral collisions.

 
Updated 6/6/17:  We modified the vertex cuts for the ZDC and the VPD to make it more effective, under the relaxed assumption that |vz|<50 cm is acceptable offline, rather than the tighter |vz|<30 cm requirement.  We also added a TOFMult>0 requirement.


Updated 6/9/17

Took a further look at 580021, split into satisfying the lastDSM requirements for the ZDC and VPD.
There is some odd behavior at low refmult, which looks like vertex confusion.  Vertices are all over the
place and there are very low refmult events with very high offline tofmult and bbc adc sum, which might be causing the big
bump at low multiplicity. The central inefficiency of the zdc is clear, as is the peripheral inefficiency of the VPD.
Don't have a great conversion of refmult into centrality, but my guess is that the VPD is inefficient to something like 60 or 70% central.
[endpoints of dN/d(refmult^{1/4}) probably something like 1.2 and 4.2, and inefficient to 2.2]