I am undertaking a review of all my instances of name changing to see if I can find some consistent policy. Of course, I take note of name variations in the notes attached to individuals but oftentimes one doesn't know if the variants are errors made by census enumerators or some other official or whether they represent the spelling used by the person him/herself. The Wares in my tree are especially problematical in this respect. It was only in about 1850ish that they all seemed to settle on 'Ware;' previously they had been variously known as Ware, Waire, Wair, Weare, Wire, Wyer, etc. But, for the most part, I don't know what spelling each of them favored or whether they were consistent throughout their own lives.
The Rosewarne/Rosewaine case is different, though, because he made a conscious decision to change his name and I know when he did it. I think I even know why he did it. He had joined the Marines as a youth and it was later discovered that he was underage and he was discharged for that reason. Since he was apparently set on a military career, he reenlisted when he was older and, at that point, he changed his name, presumably to disassociate himself from his previous aborted time in the Marine Corps.
I have another person who abruptly changed his first and middle names when he was in his thirties. The new names were entirely different from the old but at least he kept his surname intact. There was also one who was registered as Sarah Elizabeth but was always called Florence G.