I had to reboot a couple games before. Currently Fast & Furious: Seattle is doing the best. I stayed flexible with it. Really I just looked at what caused the previous version to falter, and I just adjusted those things in the next version.
Specifically for that game it had a lot to do with player posting rates, as well as an unclear direction (due to kind of a three-way tear between Cops, Crew, and Mafia).
So when I rebooted I decided to scrap the Police playing a major role (since the game was centered mostly around the underground anyways - cops didn't get much interaction). I did keep the undercover-cop player, because it seemed to be an interesting and useful dynamic.
I ended up changing the location and giving it a twist to make it different from the franchise movies, by having them stealing cars instead of trafficking drugs or stealing electronics. It really helped create that car-centered theme of the game. Moving it to a completely different environment (Seattle) from San Diego, gave it a refreshing feel - even for players who wanted to hang on to their old characters and just tweak their bios.
Appointing a couple Co-GMs made a world of difference too - even though I mostly handle everything, it at least gave a way for someone to make choices during my work days or at times when I couldn't check PMs daily. Helped maintain a better posting rate - even though the game only has about 10 players; having 3 GMs to turn to with questions I think mitigated the infrequent posting.
I'm also much more careful recruiting prospective players these days.
My Marvel game (not a reboot) I implemented a group voting system among the Co-GMs (6 of them, and myself being the seven and tiebreaker). Basically they all take a peak at an interested player's profile and previous posts to determine if they feel like the player is a good fit for the game. After a majority vote comes in, I let the new player know and they start drawing up a character sheet. Then the Co-GMs get a look and give their input for edits and approve the final draft (again by a quick yay or nay vote).
Granted that system is only really ideal for pretty large games - but it's something to consider if you're having to reboot a game. As usually (in my experience) with group games posting rate or OOC player vs player disagreements tend to cause the most failure. Careful recruiting will help a little; but life still happens.
Creating something in the game where players or characters can easily be written out or written in has helped me with the problems that better recruiting doesn't fix.
Hope those thoughts/ideas/suggestions help for future game creation/rebooting.