It's whatever split maximizes the value of the 5th runner, provided that the 5th isn't greater than any of the proceeding (and to the extent it's possible). In your example 10 > 2, hence selected.

a simple change fixed it, but annoying nonetheless. this wasn't anything complex or reliant on an external package; it's a legit error with this site's interpreter

Wrote some code in R. In RStudio the code worked flawlessly on the sample tests, yet in the test it produces dramatically different results. For example, when I run the first test case in RStudio, the first chunk correctly outputs "ijJ tipvme ibw", however here as part of the test (same code), the result is returning "ij10 209162213"