Pierre Astier has noticed a problem with an easy solution in the tests for the new SIP fitter introduced in DM-3549:

Anyway, going again through the test I realized that you had chosen order=3 and npoints=10. If understand correctly, with these values, you have zero degrees of freedom, which means that you might not be testing your error-handling setup, because the result does not depend at all on the errors. I would suggest that since you have a test routine, you carry both an over- and under-constrained test cases.