Switch to a larger instance type for your Dataflow workers to provide more CPU and memory resources.

Consider creating a temporary buffer table in Bigtable before transferring data to BigQuery, though this may add complexity.

Increase the maximum number of workers in your pipeline to distribute the workload more evenly.

Relocate your Dataflow pipeline to the us-central1 zone, which might not solve the issue and could increase latency.

Opt for creating a temporary buffer table in Cloud Spanner as an intermediate step, acknowledging potential overhead.