Ideally we should also add some direct tests for the worker CLIs. And when the network integration services do not start up successfully, we still need a proper error message... (currently just the docker exec ... pytest waits indefinitely)

Ideally we should also add some direct tests for the worker CLIs.

What exactly do you mean with that?

A new test case for the processor CLI like ocrd-dummy worker --help and ocrd-dummy worker --database mongodb://... --queue amqp://...

And when the network integration services do not start up successfully, we still need a proper error message... (currently just the docker exec ... pytest waits indefinitely)

Would it help if all docker test container logs of the processing server and processing workers are written to a file, the printed by the ocrd_core_test instance instead of yes?

Yes! That would be ideal. With that we could directly timeout 40s grep in the log file for certain events before running pytest.