If I'm reading the failed checks correctly, I got 'unlucky' with the random numbers for the test vector / test matrix.
It does not seem like a problem in the code itself, just that Cholesky is relatively unstable.
I will hard code a testcase instead.

Oh I just noticed that I was giving the univariable testcase the same y as the multivariable case, which has a bigger error term. I fixed that.
Would it still be better style to have a hardcoded testcase, than a random one?

Codecov Report

Merging #164 (acbdb42) into developer (cc228f4) will increase coverage by 0.41%.
The diff coverage is 77.41%.

@@              Coverage Diff              @@
##           developer     #164      +/-   ##
=============================================
+ Coverage      17.37%   17.79%   +0.41%     
=============================================
  Files            114      114              
  Lines           9281     9312      +31     
  Branches        1148     1150       +2     
=============================================
+ Hits            1613     1657      +44     
+ Misses          7390     7369      -21     
- Partials         278      286       +8     

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cc228f4...acbdb42. Read the comment docs.

If the test with random samples may fail, a discrete case is preferred. Otherwise a hypothetical failure in the future would lead to unnecessary confusion while the methods works perfectly fine. If a failure of this method should be impossible the test you provided seems fine.