The usual term in this context is 'violation' (of a standard). Here the item under test is apparently being exposed to test data or test transactions containing intentional violations of the standard in order to determine how it handles these non-compliant situations.

This has nothing to do with the legal sense of Verstoss.

sample refs:

[DOC]
COMMENT FORM
File Format: Microsoft Word - View as HTML
The drafting team deleted “levels of non-compliance” and added “violation severity levels” to comply with the revised Reliability Standard Development ...
www.nerc.com/docs/standards/sar/Comment_Form_Relay_Loadabil...

[PDF]
Characterizing High-Speed Optical Transmitters: Compliance Testing ...
File Format: PDF/Adobe Acrobat - View as HTML
17 Sep 2004 ... procedures and best practices for compliance testing at what is commonly ..... separately tabulate the number of standard mask violations as ...
www.home.agilent.com/agilent/redirector.jspx?action=ref&cna...

--------------------------------------------------
Note added at 17 hrs (2009-01-07 21:43:45 GMT) Post-grading
--------------------------------------------------

Post-grading afterthought:
Another interpretation of the scenario is that the unit is first tested for compliance, the results of this test are analysed, and all non-compliances (violations of the standard) are recorded. These non-compliances are then corrected by the producer or supplier, after which the unit is again tested with respect to five 'characteristic' ('exemplarisch') non-compliances selected by the customer, in order to verify that they have been corrected. This is a sort of random sample testing, since the supplier does not know in advance which non-compliances will be selected for retesting, so successful retesting implies that all of the recorded non-compliances have been corrected.
With this interpretation, you might still use 'example non-compliances' or 'example violations', but 'sample non-compliances' and 'characteristic non-compliances' would probably be better.