Now that I'm trying to write a tutorial on using the BigQuery Storage API to download query results, I've had a change of heart regarding whether to_dataframe should fall back to the tabledata.list API when there are authentication errors. Originally, I thought we should fallback on authentication errors, but now I see that it hides problems.

Someone could have downloaded the right libraries but either don't have the right permissions or haven't enabled the API, but those errors will be swallowed. Instead, downloading large tables will be much slower than expected.