Using pull requests as your code review doesn't force you to not run the code. My team uses forks of upstream repos and all merges upstream happen via a PR that is peer reviewed. It isn't uncommon for us to pull down the changes of a PR locally to "play around" with a new feature or change to existing functionality.

We have tests in place to help stop regressions, but sometimes seeing a change in action can really help to put the corresponding code in context.

Most of my team uses hub, but you can check out a pull request easily enough with git:

    git fetch <remote> +refs/pull/<pull-request-number>/head
    git checkout FETCH_HEAD