OOP objects are only 'objects in reality' by reference. Right? An actual employee at a a company is an object (but don't tell them that, they don't like it). And that object 'is' (in an odd set-theoretical, naming-relation-upon-equivalence-class-via-relation way) the OOP 'object' that instantiates "class Employee: public Human" because we expect the operations on the latter to represent real facts about the former.
I am not sure it is a good definition, but if the criterion for being a physical phenomenon is the applicability of (some future) physics to them, then all physical phenomena are objects from the OOP point of view, in this same sort of referential way. Physics is stated in laws that take mathematical forms. Any foreseeable physics will still couch its entire range of predictions in the form of a theory with clauses that identify to what the equations apply and then some methods for prediction. That seems to be the form theories take there, and anything that departs significantly from that form in the future, will probably not still be physics.
Such a law is an object, it produces the referents as sub-objects, works on them, and discards them. So an atom is obviously not an OOP 'instance of a class', in reality, but a Bohr atom, or a deBroglie atom is one, because Bohr's and deBroglie's theories are objects and they attach to a certain part of our observational space and force us to re-encode whatever we see or imagine there into a specific finite representation. This is just the same way we encode the employee as "new Employee(...)", only vague and computationally intractable. Those objects will always have the form of an OOP object, just because that is what a theory has become for us -- a list of definitions, conditions and predictions. Then the atom is the object those instances model in the same sense the real, physical employee a the object modeled by our cruder model of 'class Employee'.
[As noted elsewhere at great length I can't buy that definition, because I think we are no longer able to identify what we would and would not accept as a 'physical' theory, except to the degree it is just a form in which we want our science expressed. You can make up math that predicts anything you can imagine, so it begs the question of what is physical just to say it could be described mathematically. ('Reductivist monism' may be well defined but 'physical' is not, anymore.)]
If you are an hard-core idealist, it is a lot more obvious everything to which we can make a reference is an object. Any clear idea of an entity or process is a model (consistent or otherwise) in some category of models, and OOP is just a framing of category theory. But then, if you admit math as ideal, you have to deal with the intrinsic conflict between universality and negation (Russel's paradox).
Something cannot both be ideal and at the same time have to be coped to fit observed problems with its reality. Formalistic solutions to math's boundary-issues discard universal universality -- only some things get to be universal, and the motivations as to which things those are are from practical observation, not principle. To me these are not acceptable solutions from an idealistic point of view, because there is no good idealistic reason to seek them. The few less formalistic, idealist solutions we have found, usually deeply impoverish math, and make it a lot harder to move forward.