Thanks David!
I'm with you on it being a query limit: that's what I first thought too. But I only have 40000 rows, and each of my queries should only return at most five rows. I'm using "project" to only call three columns, and I'm using a pretty strict "where" clause.
which makes me think it's not a query limit. I am iterating through a CSV file to do "matches" with KQL against the Data Explorer API, so I'm making 40,000 requests. I increased the server timeout to the max (60 min), but it still only makes about 20,000 requests before I bombs. So what should I be writing the rate limit to?