I think we may be talking about different things. It’s one thing to trace the development of certain ideas or doctrines and notice they developed. It’s something else to decide whether they existed in the earliest canon or not. I think some ideas can be ruled out because they are rare or not consistent across canons, or we can see them developed in later literature - like the anusayas in Sarvastivada Abhidharma. This is really the main criteria I’ve found that’s reliable and makes sense to me.
Abhidharma is derivative of the early Buddhist teachings. I.e., it’s summarizing and collating existing teachings to make it easier to teach and study them. It may have been a strategy to overcome the size of the canon after it had grown too large for most people to memorize. The earliest Abhidharma is probably pre-writing, so it’s really old. The bulk of it, especially the Sarvastivada Abhidharma, is a later development. Buddhists became philosophical and began developing original theories rather than summarizing standard teachings. And we can find some of these new ideas in the sutras, like the anusayas that replaced the asavas in the Sarvastivada theory of realizing the four noble truths. This found its way into the Theravada suttas like other later ideas from northern India like the thousand-world systems and such.
I can’t say that the five khandhas are definitely from the original teachings of the Buddha. It just seems likely given that they are found in every Buddhist canon, from Theravada to Yogacara. If there was evidence of the five aggregates developing into the form that’s commonly found, I think I’d have more doubt. Instead, we see ideas forming around the five khandhas, like sakkaya, nama-rupa, etc. Khandhas look older to me as a result.
The four noble truths are a little more uncertain to me than the five aggregates, but not that much. This is because the four principles: X, the formation of X, the cessation of X, and the path to the cessation of X can be found applied to other things besides dukkha in suttas. So, then, I have to wonder if someone decided to elevate the formula applied to dukkha as the essential teaching of the Buddha. Or did people decide to use the same formula on other things based on the 4NTs? Either scenario makes sense to me, so I don’t know. SN and AN are full of variations and combinations of themes like that, so it’s not that strange. But every Buddhist tradition is adamant that the four noble truths was the Buddha’s first sermon, so I assume there’s a good reason for that. It must go back quite far if it isn’t original.
There are cases that are more certain to me. One is the ten powers of the Tathagata. I’ve written about it on the forum when I was looking at its parallels a couple years ago. Almost every version of the ten powers is different. But they are different in a consistent way. They all have a core of five powers they agree on, and then add five more powers in different ways. So, it seems that originally it was five powers, but then at some point Buddhists all decided to expand it to ten - in different ways. The way Buddhists are consistently inconsistent can creates an objective basis for a conclusion. It’s just not very often that it’s that clear.
Basically, studying parallels opens up a much wider view of early Buddhism than limiting one’s scope to only Pali sources. But it also shows a convoluted relationship between different early canons. Pali sources often agree with Dharmaguptaka parallels closely, but then they also appear to borrow from Sarvastivada sources. More and more, I view it as somewhat patchwork. There was cross canon sharing taking place at different times, I think.
. Text analysts always use word frequencies/proportions as far as I know when comparing texts. Conclusions can’t be made other than this being random chance without additional statistical rigour to back it up.