If A. Random Developer can't tell that/if her changes work before checking them in, then ... how are they to justify checking them in at all? A first alternat

If A. Random Developer can't tell that/if her changes work before checking them in, then ... how are they to justify checking them in at all? A first alternate to actually testing things might be "conformance to a model of Best Practices for changes that we think might - maybe - do minimum damage."