Some thinks that the Blessed One’s discourses can be distinguished using following characteristics.
- Gradual flow (begins with the simplest and gradually goes deeper)
- Ending with Nibbana
- No missing points in the middle (or beginning or end)
- Usage of the most suitable simile (when using similes)
- Nothing can be added or removed except expanding.
Well, that’s a start. Perhaps you might like to check out the book by Ven Brahmali and myself on this topic, The Authenticity of the Early Buddhist Texts. In general, we look for multiple independent lines of evidence.
For example, the criteria you give above mostly pertain to the coherence and meaning of a text. It is quite reasonable to assume that the Buddha spoke in a way that was coherent, and that sometimes later editions lack such coherence, either due to accident or intentional expansion. However, but itself it is not conclusive, for it is quite possible that a late text can also be coherent and meaningful.
However, if we can point to other, independent features of a text, we can be more confident. Such things might be:
- Historical context (for example, reference to certain kings or political entities)
- Use of language, terms, and ideas
- Existence of parallels in Chinese and elsewhere
And so on. When several such things all point in the same direction, we can be fairly confident that a text is late.
It is, however, much more difficult to establish that a specific text is early. Generally speaking, I think, the simplest and most reasonable hypothesis for the origin of the early Buddhist texts is that they are what they say they are. If they are not descended from the Buddha and his disciples, where do they come from? No-one has really given a serious alternative. There is strong internal evidence that the mass of texts we consider to be early do in fact derive from the time of the Buddha or shortly after. So unless something shows late features it is, I think, reasonable to assume that it is probably authentic.