After Nance left, we moved the entire setup from room C124 to C155. In this new location, several problems were seen. The first was that the data packets received were garbled. Eventually this was traced to a reversed ribbon cable. Many ribbon cables around the system are not keyed, so they can easily be inserted upside down.
The bigger problem was that the xilinx chip would crash, usually within several minutes of being loaded. Improving the grounding and shielding had no effect on this. One major problem was spotted in the process, namely that the -5.2 power supply was badly underrated, with voltage sagging to 3.1 V. A bigger power supply solved this, but not the crashing of the xilinx chip. Eventually the problem was traced to unused, unterminated driver chips, which would go into wild oscillations. Removing these chips (c22-25, c28-29) stopped the repeated crashes.