Short summary: The three HV crates have been moved to bbps1.jlab.org (same ports). The online documentation should be updated.
The new'ish dvcstlab1 portserver (or whatever it was called) was getting hit with portscans. This seemed to correlate with the HV crates getting screwed up.
I can also come up with a hand-wavy hypothesis that may explain the MWDC thresholds getting reset to 0 periodically...
It turned out that the bbps1.jlab.org address was in used by a portserver installed by the DVCS group in the right arm. When I found it, it had nothing plugged into it, but was still on the network. (This explains why the BigBite weldment's portserver seemed 'broken' -- it was fighting with the other portserver for the IP address.)
I removed the portserver from the right arm and installed it (as bbps1) into the weldment. The three HV crates and the 1 camac serial interface were moved to the bbps1 portserver (same ports). The HV GUI HVframes.conf files were updated and the GUIs verified to work.
Figure 1 is a snapshot of the portserver configuration.