Title: Features, paradoxes and amendments of perturbative non-Hermitian quantum mechanics

Abstract: Quantum mechanics of unitary systems is considered in quasi-Hermitian representation. In this framework the concept of perturbation is found counterintuitive, for three reasons. The first one is that in this formalism we are allowed to change the physical Hilbert-space norm. Thus, in a preselected Hamiltonian $H(\lambda)=H_0+\lambda\,H_1$ the size (and, hence, influence) of the perturbation cannot always be kept under a reliable control. Often, an enhanced sensitivity to perturbations is observed, for this reason, in open quantum systems. Second, even when we consider just a closed quantum system in which the influence of $H_1\neq H_1^\dagger$ is guaranteed to be small, the correct probabilistic interpretation of the system remains ambiguous, mainly due to the non-uniqueness of the physical Hilbert-space inner-product metric~$\Theta$. Third, even if we decide to ignore the ambiguity and if we pick up just any one of the eligible metrics (which reduces the scope of the theory of course), such a choice would still vary with $\lambda$. In our paper it is shown that all of these three obstacles can be circumvented via just a mild amendment of the Rayleigh-Schr\"{o}dinger perturbation-expansion approach. The flexibility of $\Theta=\Theta(\lambda)$ is shown to remain tractable while opening several new model-building horizons including the study of generic random perturbations and/or of multiple specific non-Hermitian toy models. In parallel, several paradoxes and open questions are shown to survive.