Change how we count notifications in redis

[jzbahrai]

Summary

The batch write pattern in process_ses_results:

1. process_notifications() calls notifications_dao._update_notification_statuses(updates) — bulk-writes ALL N terminal statuses to the DB atomically
2. Then loops over each notification calling update_annual_limit_and_bounce_rate()
   Inside that function, get_annual_limit_notifications_v3() is called which checks if seeding has occurred

The race condition:

1. Notification Get rid of Gov.uk Branding in UI #1 in the batch: not yet seeded → seed_data_in_redis() fires → queries DB → finds ALL N notifications already at terminal status (from step 1) → seeds Redis with all N → did_we_seed=True → skips increment ✓
2. Notification When using our copy of the notify repo we should connect it to our github actions/apps so we can automate as much testing as we can #2..N in the batch: already seeded → did_we_seed=False → increments count → but these were already counted during the seed → double count ✗

[Copilot]

Pull request overview

This PR addresses a race condition in SES batch receipt processing that could double-count annual limit notification totals in Redis when seeding occurs during a batch update.

Changes:

• Moves the annual-limit “seed state” check to run once per service after the batch DB write, then passes that seed result into per-notification counting logic.
• Updates update_annual_limit_and_bounce_rate to accept did_we_seed from the caller instead of re-checking per notification.
• Adds tests to validate that seeding during a batch prevents per-notification increments and that the seed check is performed once per service.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
app/celery/process_ses_receipts_tasks.py	Checks annual-limit seed state once per service after batch DB update and uses it to prevent double counting.
tests/app/celery/test_process_ses_receipts_tasks.py	Adds regression tests covering seeding vs incrementing behavior and ensuring one seed check per service.

Comments suppressed due to low confidence (1)

app/celery/process_ses_receipts_tasks.py:235

• This log statement uses a double-quoted string literal inside an f-string expression (aws_response_dict["message"]). In Python this form raises a SyntaxError at import time; use single quotes for the key (aws_response_dict['message']) or assign the value to a variable before formatting.

    if not is_success:
        current_app.logger.info(f"{log_prefix} Delivery failed with error: {aws_response_dict["message"]}")

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.