I think the latest answers form a neat set of points which can be considered in sequence.
@Bobbie and Melanie. To focus on what I think is most important in what you are saying, I certainly agree that when we make changes we should have sources to back up our work. However, in this case I think that for some years the situation has been stuck. While the family as a whole certainly exists, and we can therefore look at sources about the whole family, there are a bunch of Williams connected to it in Wikitree who seem to be an artifact of merges and memes.
In other words, some profiles can't really be connected to good sources. They are completely different from what we find in good sources. Wikitree can get stuck if there are demands that we must to find a sources saying that Wikitree is wrong in such cases. That would effectively be asking for evidence of no evidence. I hope that makes sense!
@John. I agree with the idea of trying to understand the history of the profiles (and therefore the intentions of past editors) in cases where there are no sources. I also try to do that. In this case, and in the case of merges generally, I sometimes find it difficult.
FWIW the strong impression I get when trying to find some kind of explanation of the intentions of past editors in this case is that (as often on Wikitree) the dates have been shifted around just to make the result of gedcom merges etc look more believable. I think we all know that this is commonly seen as good practice by many Wikitree editors and is very common. Wikitree editors do not get outraged about such changes being done without any sources. But in cases like this it only makes it more difficult to reconstruct what real people, if any, are really being referred to in our profiles.
A common type of history might be that old gedcoms somewhere on the internet gave different birth or death guesses to the same person, and somewhere on the internet, perhaps on Wikitree but probably on a gedcom which eventually came to Wikitree, these different versions of the same person became different people.
Question to everyone: in practice, at what point do we accept that there is no way to reconstruct any significant type of evidence-based genealogical reasoning from an unsourced profile, and just switch to what good published sources say? In practice at the moment we are sometimes getting stuck for years.
IMHO best practice would be first trying to search for any recent source that has questioned the solid looking scenario in Keats-Rohan and Sanders, but I do not have time to do it right now. Looking around quickly though, I could not see anything, and the "one William" scenario looks very well founded in primary sources. (There is also a grandson named William who we do not have, FWIW.)