A scan of the rawdata (using xcefdmp) revealed that there were indeed events were there was NO data from the CAEN v767 TDC... only a header (which contains the module location and event number) and EOB (which contains the module location and error bits). The EOB error bits were all off.
The frequency of these missing reference channel events, increases with the hit rate into the TDC (verified by decreasing the chamber threshold).
We verified that the reference channel was indeed making it to the TDC by fanning out the ribbon cable input to a FASTBUS TDC. The FASTBUS did not see the same problem. We also did a simple count of Level 1 Accepts and reference channel hits with a LED scaler.... indicating no problem with the VME module that distributes the reference channel signal.
We also attempted to simulate the rate from the drift chambers with a 103kHz pulser, plugged into many of the channels of the ribbon cable (along with the reference channel). We noted no missing reference channels in that simulation (VME or FASTBUS). From this.. it appears that a random rate is key to generating these missing reference channel events.
There's probably more to say on this... but it seems clear that it's not
a good idea to run like this, when we're expecting high hit rates into
the TDCs.