Here is an interesting post on how to make reverse use of GIT bisect to find fix-introducing commits instead of fault-introducing commits. (The link is broken, but the original post can be found at archive.org.) What it boils down to is this:
Since git bisect was designed to find regressions, we need to flip the meanings of “bad” and “good” in order to use it to find a fix.
It seems to me that using this would be helpful for anyone wanting to use the regression test selection (RTS) method we introduced a few years back.
Check my publications page for more details.
Update 2014-01-01: this reverse use of GIT bisect has recently been covered on stackoverflow.