we might want custom retry logic for getQueryResults

Since getQueryResults translates job failures into HTTP error response codes, we might be retrying getQueryResults more often than needed, delaying when we re-issue a query (when safe to do so). Probably will want a different set of default error codes for getQueryResults.

We'll need to be careful with this, though, as we don't want to re-issue queries when they aren't actually corresponding to failed query jobs.