The problem was occurring intermittently during the execution of a ~15 minute TestStand 3.1 sequence with LabView 7.1 calls, and it usually manifested itself as a VISA Write timeout error. I was not able to reproduce the error manually in a reliable fashion by executing our drivers manually, but I found that a VISA Clear command would re-instate instrument communication without having to turn the instrument off manually. Increasing VISA timeout values had no effect (just increased the wait before the error message would appear). Synchronous vs. asynchronous had no effect either, although the AmRel documentation specifically states that it supports asynchronous communication (as I believe all 488.1 instruments must?).
We contacted AmRel about our problem, and their response was:
"There are no known communication issues with National Insturments [sic] VISA. However, the PD8-2D (488.1) firmware does not have command flow control. If a fast computer is use [sic], e.g., P4 computer we suggest adding about 50~70mSec after each read or write command."
I made this change to our drivers, and the power supply now appears to be communicating just fine in our sequences. I believe that the instrument was slow to respond to VISA Read requests, and occasionally data would be "left behind"; thus, the bus would be tied up at the next VISA Write occurrence. I'm not excited about inserting a "guesstimate" delay into the drivers, particularly one prone to variation from one execution to the next, but I certainly prefer it to the prospect of having to manually reset the power supply during production testing!
As a sidenote, this is basically how the drivers (written in LV 4) that AmRel ships with the power supply are constructed. We had re-written the drivers to eliminate the delays, and they worked fine with our previous test station configuration. It was only upon switching to a new test station configuration (which included a faster computer, LV and TS version upgrades, and new VISA drivers) that the problem appeared.
In any case, problem solved. Thanks for the advice!