We'll do a similar protocol as other bulk import / migration processes (like Notifications):

1. Pause listener / ongoing sync
2. Run import job
3. Resume listener

This ensures that disabled users are not re-enabled by the import. If we didn't do this, then this scenario could happen:

1. Import job starts with exported state of user 1 as enabled
2. User 1 is disabled
3. Listener disables user 1
4. Import job imports user 1 but prior to disabled state

This would mean user 1 would have access even though they are disabled.

The protocol above prevents this. It requires a toggle which, when off, does NOT consume messages, and when on, resumes consumption from when the toggle was last turned off.

Q&A

How frequent are user events?

About 5-8k / day, see splunk query 

`umbbreadcrumb(environment=prod)` | search destination="topic://VirtualTopic.canonical.user*" | timechart count by op