Class NotificationJobEnqueuer

Remarks

The following are true:

1. There is a maximum of MAXIMUM_RECIPIENTS_COUNT recipients per trigger request.
2. Our notification provider throws if we use the same idempotency key, but the body changes.
3. We want to be able to query for template variables in jobs so we're getting the most up-to-date values.

Taken together, we need to ensure each job only contains one chunk of recipients so it either succeeds or fails and retries on its own. If we moved chunking to the NotificationClient.trigger method, for example, the following could happen:

1. A job with >MAXIMUM_RECIPIENTS_COUNT recipients runs.
2. The first chunk succeeds, but the second fails because of a transient issue.
3. The job retries, but a first batch template variable changes in the meantime.

Now the job will fail indefinitely and the later batches won't get their notifications.

Even if you're sure you won't have more >MAXIMUM_RECIPIENTS_COUNT recipients, the method enforces other best practices, like setting expiresAt on enqueue instead of calculating it in your job on each run. Doing this usually means it's always in the future and doesn't help prevent stale notifications.

So please, use this method if not for customers, then to save fellow engineers time debugging!