"git rebase -i" is essentially how the entire Linux kernel process works - and IMO it works pretty beautifully. The only real downside I've seen is that it can lead to developers submitting code which breaks bisection due to not testing each individual rebased commit - in the kernel, this is somewhat alleviated by the Intel 0day bot which gives us automated build failure feedback, but it's still an issue. The benefits of a nice clean changelog though should not be underestimated.