I recall seeing a study a few years back that showed direct correlation
between packet loss and temperature on DEC Gigaswitches; does anyone else
remember this?

I posted a correlation between packet loss at temperature at MAE-East (or
Boone more precisely) to NANOG a couple of years ago. Executive summary:

* A lot, but not all, of the correlation was put down to the fact that
  high traffic levels cause both high temperatures and high packet loss.
  This explained by no means all of the problem, as pointing a fan
  unit at our router, the concentrator etc. reduced packet loss.

* Cisco 7010's and CX-FIP cards degraded under excessive temperature
  (above about 35 degrees C).

* MFS (as then)'s FDDI concentrators, and possibly Gigaswitches degraded
  under similar conditions

* Surprise surprise, this didn't explain all packet loss at the MAE.

But yes, it did make a *real* difference.