I'm having a lot of trouble understanding bisect results from the refactor. This is a meta-bug for weird results; I'm blocking it on the sub-bugs. To keep the feature launched we need to be able to triage these bisects to the point where we can:
1) File clear bugs on problems in bisect itself
2) File clear bugs on external failures (infra issues, test issues) to the point that test authors, troopers, infra labs, etc can understand the problems and address them easily.