What Now? Handling Errors in Large Systems / hacker news