The effects were figured out to the best understanding of the coder doing the work. But he's working on code he didn't write, and the code is difficult to understand and has no comments. In such cases, you don't always know if you've understood the code properly until you write tests to check its functionality.

This is doubly true because the code can simply have errors that make it operate in a way that wasn't even intended. We already found a case of this, where a manually written sqrt function contained a bug. So testing and third-party review of both the code and the test results is important. You can't just stare at the code and be sure you understand it entirely. That kind of "ivory tower" programming is a recipe for disaster.

As to witness voting, I vote for witnesses on different criteria. In abit's case, despite the fact that he devotes most of his time to BitShares, I vote for him because he has more knowledge of the core code than most of the other witnesses. This can be very valuable at times. But note that my vote alone doesn't keep him in the top 20. He's fallen out at various times, and I think he's only on the edge of being in now.