I've twice in my career found reasons that git being (officially; I have no interest in dealing with another implementation with its own missing features and distinct bugs) a library instead of a messy ball of scripts and disparate binaries, would have saved me tons and tons of time. You can look at the stories of how Github was designed and built, or look at the architectures of other similar software, and see folks struggling with the same issue. You'll run into frustration on this front pretty much instantly if you try to build tooling around Git, which turns out to be such a useful thing to do that I've ended up doing it twice in ~15 years without particularly looking for reasons to.
(While we're at it, how about some kind of an officially-blessed lib-rsync with a really pleasant API?)