I'm not an expert on this, but I think the surface code is so popular because it's 2-dimensionally local and has great parameters and speed of error correction. But people certainly work on other codes too, and if someone invents something that's unequivocally better than the surface code---or better for what the experimentalists are building now---I guarantee you they'll take notice! They don't always care about finer points of complexity theory, but this they do care about.