There is no experimental evidence that prevents the universe from being described by wildly nonlinear transformation equations.
What are you even talking about? In principle, you can parameterise space and time using
any coordinates you want. The only restriction is that coordinates should faithfully represent the topology of space-time.
If you understand that the Lorentz transformation is supposed to represent a
symmetry of existing physical laws, then you're plain wrong. The Standard Model of particle physics successfully describes all non-gravitational interactions seen in nature, and is well verified experimentally. General relativity is also quite well verified. Between them, both theories cover the currently accessible experimental domain. Both have clear mathematical definitions and there is no ambiguity concerning the symmetries they possess. You are not entitled to just make up any claims you want about physics and present them as fact.
they are just arm-waving to justify their pretense that their argument is geometrical.
No they're not. They're requiring that the laws of physics remain the same at every point in space and time - ie. they should be invariant under coordinate translations. Mathematically, we require that the Lorentz and translation groups commute. In other words if we choose to apply a translation to a system followed by a Lorentz boost, there should be a way of first applying a boost followed by a translation which produces the same result. In mathematical terms, if $$x^{\prime} \,:\, \mathbb{R}^{4} \,\rightarrow\, \mathbb{R}^{4} \,:\, x \,\mapsto\, x^{\prime}(x)$$ is your boost, then for all $$a \,\in\, \mathbb{R}^{4}$$ there must exist $$b \,\in\, \mathbb{R}^{4}$$ such that
$$x^{\prime}(x) \,+\, b \,=\, x^{\prime}(x \,+\, a)$$
for all $$x \,\in\, \mathbb{R}$$. Taking the partial derivative of both sides produces
$$\frac{\partial x^{\prime}}{\partial x}(x) \,=\, \frac{\partial x^{\prime}}{\partial x}(x \,+\, a) \;.$$
This has to hold for all $$x$$ and $$a$$, so $$\frac{\partial x^{\prime}}{\partial x}$$ must be a constant, and the boost $$x \,\mapsto\, x^{\prime}(x)$$ must be a linear or affine transformation. Anything else is inevitably going to rely on the existence of a privileged coordinate origin.
I don't call that "arm-waving".
But "The aim of
The Quintessence of Axiomatized Special Relativity is to remove from Einstein's relativity theory everything that is confused, unnecessary and not amenable to experimental verification."
Then you're trying to fix what isn't broken.
You may be confused about relativity. That doesn't mean the rest of us are.