It probably didn't sell because it wasn't very good. So you re-balance it later and now it doesn't suck. Like, fundamentally keeping the "best" and "worst" models/armies/strategies from stagnating keeps the game interesting (and drives more sales... so depends how you look at it).
I don't think they've every been super good at balancing though, and that at least is a fair criticism - albeit a hard task given how time consuming playtesting is to get data.