It seems that the current Google BigQuery writer firstly writes all the features into a temporary CSV file then loads them into the destination BigQuery table. I thinks it's a good mechanism if you load many features into a BigQuery table at once, but it may not be good if you need to write a single (or a few) records into the table frequently (e.g. write 1 feature per a minute), since BigQuery has limits the maximum rate of incoming requests.
Quotas and limits: https://cloud.google.com/bigquery/quotas
However, there is another option - Streaming data into BigQuery in order to perform frequently writing features into BigQuery table avoiding the limits above.
Streaming data into BigQuery: https://cloud.google.com/bigquery/streaming-data-into-bigquery
It would be wonderful if the Google BigQuery writer could also support Streaming data into BigQuery optionally or you could add a new transformer (e.g. GoogleBigQueryStreamer) to the package.