What concerns me is that we might find some FA candidates in there. Can we think of an algorithm? DGG
You could narrow it down by looking at all deleted articles over a certain size which don't have "copyvio" or "copyright" in the deletion reason. You should be able to narrow it down a fair bit that way - you then have to plough through the articles manually. A deleted article that might be almost an FAC is quite likely not formatted correctly yet, so you won't be able to automatically search for links and refs etc.