We have an easy route to relieve ‘tension’ on the word “good” by adding an extra degree of freedom to it (what would otherwise a contradiction is no longer once I create multiple “kinds of goods”.
This gives you too many degrees of freedom - everything is good for some purpose for some kind of being, and we now need to prevent this from being abused by finding a principled reason for limiting whaat kinds of good are important in ethical normative judgments (“sure, you’re ‘good’ in some sense, but I will judge you”).
Seems like some form of “good (simpliciter)” can build in enough flexibility to account for all of the specific kinds of good (and any ‘global’ things too). When someone talks about “good for X” they are restricting our focus to a subset of the overall logic that is relevant for X, which could be useful since it’s a much simpler concept to worry about.
To claim there is no “good simpliciter” is to say that you can partition the logic of “good simpliciter” into the relevant subclasses (then by occams razor, the global good is not conceptually necessary).