You’ve premised your argument on there being a need to manipulate an array that is located (at a user-visible level) over two different physical locations. If you want to argue for the relevance of your conclusion you need to first argue for the relevance of your premise. Why does the user need to be concerned with a data structure that has been split into chunks? Could this concern be relegated to infrastructure? If so, then we don’t have a language problem. The user can simply deal with the array as if it is a single contiguous entity.

Tersely repeating myself: The main problem isn't imperative programming. The main problem is programmers wanting to "relegate concerns to infrastructure".

Programmers have been relegating concerns to infrastructure ever since they stopped programming in binary and stopped manually allocating their memory.

And how's that going? TNSTAAFL.

Err... I think, first off, that we aren't talking about an array, we're talking about some sort of list - an abstraction that maybe implements the same interface as an array. Second, it's not literally possible to manipulate the array on machine B from machine A... so what is this argument about? About whether programming languages should provide an illusion that multiple machines are one machine? About which entity should provide such an illusion (e.g. programming language vs library vs something else?)