It's important to us to test things servers as thoroughly and completely as we can. When we test things, we're actively looking for failures. We'd much rather have a component fail here, where we can replace it, and test its replacement, than have something happen at a customer. It saves everyone the trouble of having to send parts around, it saves the customer headache, and it makes things flow much more smoothly on our end. We run our diagnostics for multiple iterations, because it is possible to have something pass when it really shouldn't. Some things will fail after they've been on and warmed up for a while, and if an item fails just one test cycle, it gets replaced.