Spark stateful stream processing: joining stream with a cache

Question

Assume we have a situation as follows:

1. Unprocessed stream A is being written to S3
2. There are already processed records B stored in S3
3. Processi