We can define an "electrical device" as a device that works with electricity etc. Suppose that the theory of electromagnetism would be proved wrong in the future. Would then make sense to speak about "electrical devices"?
There are two philosophical points worth getting into here...
First, Wittgenstein would point out that when we 'define', we are creating a class designator (my terminology): e.g., creating the definition of an 'electrical device' doesn't define an object, per se, but defines the set of all objects that fall within our constraints. Further he'd note that all such class designators are 'family resemblances' other than identities. In other words, if you imagine a photo of a large extended family, with the grandparents in the center, and their children, grandchildren, cousins, nephews and such spread out to the left and right according to how closely related they are to the central figures, then as we scan across the picture we can see the family resemblance from one person to the next, but if we look at the leftmost and right-most people, they may look nothing at all alike. This allows concepts to drift — some members drop out of the picture, new members are added, new snapshots get taken — and eventually the original 'family resemblance' might be lost entirely. If electromagnetic theory were proven 'wrong' in the future, all that would happen is that we would start excluding devices that worked by the old principles and including new equivalent devices that worked by whatever new theory was developed; maybe we'd eventually change the name of the class designator, maybe not.
Second, time-conditions play hell with definitions, because definitions are (intuitively) atemporal. When we say something is an 'electrical device' we inherently presume that it was always an electrical device, and always will be. Nelson Goodman pointed out this problem back in the 1950s by inventing two new definitions: grue and bleen. Grue is the color of an object that is blue before some arbitrary but fixed date, and green thereafter; bleen is green before that same arbitrary but fixed date, and and then blue. His point is that it is logically impossible to tell whether an object is green or grue (or bleen or blue). If we buy a brilliant blue sapphire, how could we know whether we will wake up some morning and discover it's now a brilliant green sapphire, because it was actually (always) bleen, not blue. Adding time-dependent properties to our definitions really mucks with our capacity for induction.
Best not to head into this philosophical darkness. You are likely to be eaten by a grue.
Newton's theory of gravity has been proved woefully broken. It has been replaced by Einstein's theory of gravity, also known as General Relativity.
Similarly, electricity and magnetism did not change their names when Maxwell developed his unified theory of electromagnetism, nor when Dirac developed a quantum reformulation or when the quantum field theorists redefined everything.
Yet another reformulation, such as a string theoretical one, will still not change the observed phenomena, nor what we call them. Only the underlying concepts invoked by the name may change.
In this sense, yes definitions are temporary. But not because we define a different thing, only because our understanding of the same old thing's nature changes.
It seems to me that you question derives from the following reasoning
(1) one cannot define what does not exist (2) therefore, a defnition * presupposes* the existence of the definiendum (3) therefore, when the definiendum is proved not to exist, the definition looses its meaning.
I think premise (1) is questionable. However, I think it is correct to say that a definition presupposes the possibility of what is defined.
A definition, as such, does not require the existence of the definiendum to be meaningfull ( to have a sense). Witness : after having defined something, you still have to prove that it actually exists.
Let me define a "marswalker".
marswalker : a human person having walked on Mars before the 1st of january 2020.
The defining expression has a sense, you can understand it; you grasp a concept while you hear it. So " marswalker" is meaningful.
Since there is no person satisfying the definition, the denotiion of the terme " marswalker" is the empty set ( and the empty set is not nothing, it exists).
- Also, note that a definition is a biconditionnal ( not a cetegorical statement) :
for all x [ ( x is a marwalker ) if and only if ( x is a human person such that...) ]
that is ( using the biconditionnal's definition)
for all x
(1) if x is a marswalker, then x is a human person such that ...
(2) if x is a human person such that ... then x is a marswalker.
As you can see , it nowhere says or implies that there are actually " marswalkers".
Note : here I use the distinction drawn by Frege between sense and denotation.