Not sure $subj is worth doing, here are my thoughts:

We would need to have telemetry at the latest version in the whole process of bisecting, and chrome be built at revisions that the bisect chooses, I thought it works this way.

In issue 661939 we found that we probably cannot pinpoint a single regression, but if we could, a working bisect would have helped (see comment 23 there). Priority of this - I don't know.