What happened?

BQIO was lazily creating streams, which means that all our appends to a given stream would be appended at index 0.
This meant that for any second append to a stream, we would get an exception from BQ, and be forced to retry our work. This results in bad performance by itself, but combined with #26521 resulted in data loss for beam versions 2.44 - 2.47 inclusive.

Issue Priority

Priority: 1 (data loss / total loss of function)

Issue Components

• Component: Python SDK
• Component: Java SDK
• Component: Go SDK
• Component: Typescript SDK
• Component: IO connector
• Component: Beam examples
• Component: Beam playground
• Component: Beam katas
• Component: Website
• Component: Spark Runner
• Component: Flink Runner
• Component: Samza Runner
• Component: Twister2 Runner
• Component: Hazelcast Jet Runner
• Component: Google Cloud Dataflow Runner