The test used a histogram which generates different results in parallel.
Also we got failures due to minor chart differences. Making the test
more robust and hence more reliable by change the filter it tests -- now
we test the "Extract Component" filter instead. It still tests the same
domain and its UI.