Open
Description
Describe the bug
replaces #622 with a more specific issue. Since #1598 the only numba related exclusions I am aware of is ACF #625
and this one. I thought #1598 might also fix HC2, but it seems a different issue. Core of it is some numpy functions work differently if numba is turned on than when it is turned off. For example, #1598 was caused by np.argsort treating ties differently gibing different results on rare cases with Catch22. I would guess this fail is also caused by Catch22 and numba/numpy interaction, but it will require some trial and error to find it.
Steps/Code to reproduce the bug
disable numba, run results recreation
Expected results
results should (probably?) be the same with both
Actual results
test fails
Versions
No response