It can cost a lot to run a SELECT * FROM my_table query, and yet from an API perspective, this query is completely unnecessary, as one can list the rows from the table directly for free with tabledata.list. I propose

df = pandas_gbq.read_gbq("my-project.my_dataset.my_table")

which will download a table directly (using bigquery_client.list_rows().to_dataframe()). Since all queries must contain some kind of whitespace and table IDs cannot contain any whitespace, we can use that to disambiguate query versus table ID.