The nightly build fails most of the time currently, since either waiting for the data generation to finish, or the Optimize import are not properly waiting to be finished before continuing to assert the results.
This means most of the time the imported values are inconsistent with previous runs and not the same as the generated data in the database, making the assertions fail.
A simple bandaid fix is adding an additional sleep of 1 minute before the assertion, since this is apparently enough time for Optimize to finish importing everything.
Also see the following debug test run on stage with logs on the StatusWebSocket changes on import (and the 1 minute sleep):
The issue is probably connected with the failing import performance test on large static dataset (
There, similar behavior is observed, that the import status is set to false even though he didn't finish importing, which might mean the issue lies in the importing process, rather than the data generation...
(though the dataset seems to also be an issue, since it works with the old one)