[matheist]

Semi-serious answer: Shapley value (https://en.wikipedia.org/wiki/Shapley_value) would provide a "fair" division.

Unfortunately it wouldn't be easy to actually calculate. The inputs to the formula are "how much of a donation would we have received counterfactually if the team had only included this subset of contributors" for all possible subsets of contributors. Though if it's a small number of contributors then you only have a 2^(small number) of counterfactuals to pull out of nowhere.

[MrLeap]

Here's a silly idea: take a sample of lines of code, for each line in the sample, remove the line. See how many tests fail as a result of the line being removed. shares++ to all users that come back with git blame entries on that line * how many tests fail when it's removed.

then normalize so all the users add up to 1, use that as your coefficient on disbursements.

Of course it's gameable, hopefully there's a BDFL to guard against test/commit stuffing or other bad faith things.

[ttymck]

This seems like it mostly tests whose code has the most test coverage, not which lines of code are most important?