Question
Main question:
Let $\| \cdot \|$ be a norm on a finite-dimensional real vector space $V$. If $f : V \to V$ is a function satisfying $$ \| f(x-y)\| = \|f(x) - f(y)\| $$ for all $x, y \in V$, does it follow that $f$ is additive? I.e., does it follow that $f(x + y) = f(x) + f(y)$ for all $x, y$?
Follow-up:
Does the answer change depending on the choice of $\| \cdot \|$? In particular, what if $\| \cdot \|$ is an inner product norm?
Background
Last week, user C.F.G. asked a very similar question, to which the answer was "no for trivial reasons": they asked whether the function $f$ was necessarily linear, but clearly all additive functions satisfy the condition, and there are non-linear additive functions. User Charlie Cunningham pointed out in the comments that the question is actually interesting if you remove the trivial reasons. Because the question had been answered, I tried to get an answer to the non-trivial question myself. The original question included the requirement that the vector space be finite-dimensional; I removed this requirement because it erroneously struck me as an irrelevant restriction (I thought you could just take the subspace containing $x, y, f(x), f(y)$ to get a finite-dimensional $V$). (The functional relation in those questions has a different form, but this is equivalent the formulation above, as pointed out by user Omnomnomnom in comments.)
However, we then got a negative answer to my question -- an example of a non-additive $f$ satisfying the condition -- where the infinite-dimensionality of $V$ was crucial. This has left us with the tantalizing option that the finite-dimensionality of $V$ was crucial to the resolution of the question. I did not want to ask yet another very similar question here, but I also did not want to edit my question, as it had gotten a correct answer that did not deserve to be made irrelevant. Maybe third time's the charm?
The claim does not hold. The classical counterexample mentioned in the linked answer by GEdgar works.
Let $V=\mathbb R^2$ supplied with the max-norm. Define $f$ by $$ f(x) = (x_1, |x_1|). $$ Then $$ \|f(x-y)\| = |x_1-y_1| $$ and $$ \|f(x) - f(y)\| = \max (|x_1-y_1|, \ \big||x_1|-|y_1|\big|) = |x_1-y_1|, $$ so the condition is satisfied but $f$ is not additive.