This always happens on the same node. If I SSH into that node app folder, I
see that there is a /1/, but not a /0/.
Why is it looking for the wrong folder? This is stage 16/19, so it isn't
like it bombs from the get-go - that executor has done many previous tasks.
I can't figure out how to troubleshoot any further - the spark job never
bombs, that one task just keeps running...