When bigquery read does a query it sets the mode to batch. A batch query can be very slow to schedule as it batches it with other queries. However it doesn't use batch quota which is better for some cases. However, in some cases a fast query is better (especially in timed tests). It would be a good idea to have a configuration to the bigquery source to set this per-read.

Imported from Jira BEAM-2817. Original Jira may contain additional context.
Reported by: laraschmidt.