I know everyone feels rushed, but I do think that in open source especially, git commit messages work best as lab notebook entries and not just explanations of what you did.

Nire Bryce

where you can, give context, assumptions, dead ends, etc.

not just because people need to come up to speed, but because 6 years later, the project can partially reconstruct the rationale behind a decision that's currently under scrutiny.

commit diffs already show what you did. summarize, sure, but not only that.