That adds another layer of complexity as well, because it's entirely possible to have more than one execution happening simultaneously, since we have tests running in PRs. One PR could start run one, get halfway through it, then another PR starts and resets the org, potentially erasing data that the first run was operating on