Python profiling: Cumtime - sum of subfunctions larger than parent function?

Question

In related questions, people ask about the sum of time spent in subfunctions being lower than the cumtime in a parent function, and the obvious answer is, the time was spent in the function itself.

However, for me, it is the other way round: The way I am interpreting these results, the total time spent in subfunctions is larger than the cumultative time spent in the parent and subfunctions. Obviously, this cannot really be the case - So I am either misinterpreting the results, or have convoluted code that breaks the cProfile tool? I am not using multiprocessing, or at least, not knowingly (num_cores=1)

See the screenshot below:

I used the following code to get this:

pr = cProfile.Profile()
pr.enable()
test_univariate()
pr.disable()
pr.dump_stats("univariate.stats")

and visualized the results with snakeviz.

Also, the percentage in the function information seems slightly off, as well...

Any idea where to start? Happy to provide more information.