Hida theory is a vast domain of research. I am assuming that that you are in the simplest and oldest setting: Hida theory for ordinary eigencuspforms for the group $\operatorname{GL}_2$ over $\mathbb Q$ (the arguments I am giving would apply with little modification to $\operatorname{GL}_2$ over a totally real field $F$ after an application of the Jacquet-Langlands correspondence).

In that setting the Hida-Hecke algebra $\mathbf T$ injects in the endomorphism ring of a module which is free of finite rank over $\Lambda$ (the usual Hida algebra of diamond operators), namely the ordinary part of the inverse limit on the level at $p$ of the first cohomology group of the modular curves. So $\mathbf T$ is finite, torsion-free as $\Lambda$-module and so the ring-extension $\mathbf T/P$ of $\Lambda$ for $P$ a minimal prime of $\mathbf T$ is integral.

Hence the Zariski-density of arithmetic primes in $\mathbf T/P$ reduces to the density of primes below them in $\Lambda$ which are, by definition, the principal ideals $(\gamma-\chi(\gamma)\gamma^{k-2})$ for $\chi$ a finite order character and $\gamma$ a topological generator of $\Gamma=1+p\mathbb Z_p$. These are dense in $\Lambda=\mathbb Z_{p}[[\Gamma]]$.

These arguments carry over for many reductive groups (symplectic groups, unitary groups, units in quaternion algebras...) over totally real fields but it is typically not the case that arithmetic points are dense in the spectrum of the Hida-Hecke algebra for a reductive group over a number field which is not totally real. This is already false, for instance, for $\operatorname{GL}_2$ over a quadratic imaginary field $K$.

Regarding your question on the interpolation of the $\mathcal L$ invariant of the adjoint representation, this is a very subtle result, especially in the generality you consider here. It requires first deep modularity results of Wiles, Taylor-Wiles, Skinner-Wiles and Kisin on deformations of Galois representations with coefficients in a finite extension of $\mathbb Q_p$ (so with infinite residue field). These results establish the vanishing of some Selmer group for the adjoint representation and this in turns lead to an explicit computation of the $\mathcal L$ invariant which shows that it extends to an analytic function on the set of height-one primes of the universal Hecke algebra.

Finally, regarding your reference request, I like *Control of nearly ordinary Hecke algebras* (H.Hida) and *Residually reducible representations and modular forms* (C.Skinner, A.Wiles) Inst. Hautes Études Sci. Publ. Math. No. 89 (1999), 5--126 (2000) myself, but there are dozens of excellent references (none of which make essential use of the notion of motives, so don't get scared by that word and just mentally translate it to geometric Galois representations or compatible systems of geometric Galois representations whenever it appears).