I tend to hack whatever I'm preparing to run most recently if I'm not doing something more or less my own. My FitD projects all started when I was knee deep in preparing to run Blades for the first time--it's the point when I'm most comfortable with a system and what makes it tick, I very quickly learn to correct various preconceptions I had once I actually start running it, and I haven't had time to steep in things that don't work out for me longterm. Altogether it means I'm primed to think about why mechanics do what they do but I don't fixate on fixing things and can instead focus on poking at the various bits to see if I need them for what I'm doing, and if not whether I can just rip them out or if I have to bother to put something back in their place. If a project doesn't build enough momentum or identity by the time I become more familiar with the ins, outs, pros and cons of running a system, the hack is probably dead unless I port it to my new darling or to something less directly cribbed from existing work--I just can't separate the game I'm trying to make from the (original source) game I've been playing at that point.