I've got to admit that this was the first time I've actually used glGetError() to track stuff down, but once I realized it holds the last error, I kept running it through my initialization code until I found the method that does it. Calling glGetError() after each gl* call is expensive on OES implementations, so if I do that, I'll have to setup a bunch of #ifdef preprocessors so that I can run my builds w/o being bogged down.
As far as glDisable(GL_TEXTURE_2D), I've been checking that, but this where I think the issue could lie because I actually don't call commonly used glEnable/glDisable states directly. Instead, I have my own method that holds a synchronization variable to determine whether the state is already enabled before calling glEnable(), and vice-versa. This meant to cull unnecessary calls for OES builds. I think something could be wrong with glBindTexture because I modified one of my wrapper methods recently for that particular call. I'll also have to check glActiveTexture's wrapper function...
My models also don't use just textures directly, they use a Material object that is configured to work with the corresponding config of my uber shader built for that model to render. My Material class has some logic when uploading texture types that don't exist across each mesh in my model (uber shaders work at the per-model level for efficiency). So, if some parts of my model uses normal mapping, but others don't, instead of loading two shaders, I just load one with normal mapping, and for materials that don't have normal maps associated with it, it'll send a default 2x2 normal map. If I'm using diffuse maps, then they default meshes without ambient maps to a white texture.
These textured vertices aren't lit, however, and I know it's going through the lighting shader too, so I think it's some sort of GL state issues like you've all pointed out.