There's no way to identify file that has a couple of bytes. But in many cases 100 is enough. Do you have any idea, by your knowledge / experience, where is the lower boundary for identifiable file size?
I don't call TrID for files under 16 bytes, but, for permormance reasons, it's important for me to make this limit as high as possible - without loosing correctness.
I could look for the shortest signature in your definitions, but this doesn't give me certainty that in future there'll be no shorter one, so it's rather generic, not technical question.