During PR review, we should ideally be able to get a preview of the reexecution benchmark dashboard to see how the PR will affect the dashboard. This would have prevented #4396 but more importantly, we shouldn't be testing dashboard changes via pushing to master
.