That is a partial workaround but that filter would then provide all the individuals who previously had the new place name and so they do not need sync'ing again with Ancestry.
The situation is something like this: If a hint provides a place "Here Upon There" but its official name is "Here-Upon-There" (i.e. with the hyphens) then it is easy in RM to change the non-hyphened version to the hyphened version (I normally just merge the two versions with the hyphened version remaining). However, if there are already 500 instances of the hyphened version in RM (and in Ancestry) then filtering on the hyphened version in TreeShare would still provide me with 501 instances for me to plough my way through to find the one instance that it is wrong in Ancestry.
The reason for using standardised places is so that it is easier to do statistical queries in RM and it helps when using a atlas against timelines.
I have not found an instance when Ancestry changes place names back and certainly not without me knowing about it. Obviously if there is a hint then it will use whatever was on the original record. Even then you have to individually tell Ancestry (when online) to overwrite a place name (or any other fact) with the data from the hint. Even for new facts, which have a non-standardised place name, I can cope with that by simply using the above system to bring the fact and sources into RM, clean it in RM and then sync it back to Ancestry.