This is not about whether or not getters/setters are wrong. I understand its impact to encapsulation and that question has been raised here and SO several times already. I also do not want to just ask for an example of how to change object state as all the examples I've seen are so naive that they're only practical for simple objects. I do want to ask for an example on more specific terms so that I can make the rubber meet the road in a very practical way concerning this concept.
How can an object's state can be changed when one of its data objects cannot be initialized upon construction of the parent object to begin with?
Even more specifically, I have this situation where I am defining a domain entity for an API that is representative of a data entity in a third-party system whose data is stored in files. (For this question I'm using domain/data entity to distinguish the difference between the two). This is a requirement to ensure loose coupling between layers and the fact that though the entities are representative of each other, they are not one-to-one representations, more representative in principle. The client code will be able to create these domain entities and commit an internal transaction designed to coordinate the write to the files. One of the domain entity's properties is a Handle object that contains the third-party data entity's handle ID so that any later changes to the data entity in the file can be easily and quickly coordinated. This handle ID can't be set to the domain entity until it captures it in the data access layer on the first write.
In a nutshell
So in a scenario where the object cannot obtain all its state on construction, but get the rest at a different point in the program (and maybe even a totally different application layer), how should this look without breaking principles of encapsulation and Tell-Don't-Ask?
Btw... I'm using c#, so any comparable example along these lines would help me better relate.