When sending TallySummary messages, set the message key to the org_id so that messages with the same org_id will be sent to the same topic partition, and will be processed in order by the same consumer. This would prevent remittance for a single org from getting updated in parallel.

Investigate if there is something that we can do with logging to help with verification.

Testing

• Deploy into EE with multiple partitions
• Run an hourly re-tally on all orgs that spans multiple hours.
• Verification:

• • The log messages for a single org should be processed by the same consumer (check thread)
  • Locally we can use a Kafka UI tool to verify summary messages for a single org are all in the same topic partition.