In SAP, the extractor is used by two systems: BW and Qlik. To test if that was a problem, a new system for an extractor to be used exclusively by Qlik was created. The BW system did not use Qlik's system for the extractor.
Doing this, the extraction worked.
On the other hand, for full and delta loads, this raises the question: If you utilise BW, can Qlik Replicate do full and delta loads in a standard way (without having to create Z extractor)?
Errors and Warnings to watch for:
]W: Existing extraction job found for Extractor Name. Will not send second request to delete table entries and start extraction job.
]E: An error occurred unloading dataset: Extractor Name java.lang.NullPointerException
There is a known limitation on this, as the extractor should be run by Qlik Replicate only. If you use BW and Qlik at the same time, there is a chance of missing records unless they are not crossing each other.