As part of #298, @ruebot discovered that the problem of image extraction scale was resolved in part by a more recent version of Spark as well as configuration settings. To do a large-scale image extraction job, some additional flags might be required when running spark-shell.

From #298:

Yeah, we might add a cautionary note to this section about file systems, and flags. I can help flesh that out when the time comes.
-@ruebot