When tracking down a memory leak seen on build bot, I try to narrow it down to one layout test that causes the leak, by repeatedly invoking run-webkit-tests --leaks with smaller sets of tests. However, there is no easy way to specify a continuous range of tests, and there is no way to automate the process. Building "range" support into run-webkit-tests and adding "bisect" functionality either in run-webkit-tests or in a wrapper script will help speed up the process.
Definitely similar to Bug 16873. Would be nice to be able to define what "failure" is in this case (whether it's a leak or a failing test).
<rdar://problem/9509302>