Ok. Looks like I solved my own problem. As I looked at the code, I
realized that my selection was grabbing 2 hectare polygons instead of
1/4 hectare ones. In my dataset, it was the difference between
grabbing 75% of the polygons vs. 13% of the polygons.
So, as I monitored the script before it crashed, I noticed that it was
soaking up most all of the system resources. Hence the unexpected
system error.
As soon as I changed the threshold to 1/4 ha, everything was happy.
humbled....