@januz had a very very wide dataset that REDCap had problems producing the CSV for. But the JSON was well-formed. The changes made in #335 work now on his machine, but we'd like to include a similar dictionary to detect any regression errors in the future.

I'll probably

• simulate it now with a big loop to spit out the csv dictionary, and then
• throw in some flavor from his real dictionary (that shouldn't be public)