Let A be an abelian variety, and G ⊂ Aut(A) a finite group acting freely in codimension two. We discuss whether the singular quotient A/G admits a resolution that is a Calabi-Yau manifold. While Oguiso constructed two examples in dimension 3 [32], we show that there are none in dimension 4. We also classify up to isogeny the possible abelian varieties A in arbitrary dimension.Remark 2.3. Note that if g ∈ Aut(A) admits a fixed point, then g contains no translation, so g and its matrix M (g) have the same order.Junior elements play a key role in the study of finite quotient singularities, as the following theorem emphasizes.Theorem 2.4. [18] Let C n /G be a finite Gorenstein quotient singularity, and suppose that it has a minimal model X. Then there is a natural one-to-one correspondence between conjugacy classes of junior elements in G and prime exceptional divisors in X.Remark 2.5. Note that such a minimal model X always exists as a relative minimal model of a resolution X → C n /G, by [22, 1.30.6].Quotient singularities are Q-factorial, so they can not be resolved by small birational morphisms. This yields a simple corollary of the theorem.Corollary 2.6. [18] Let C n /G be a finite Gorenstein quotient singularity, with G acting freely in codimension 1. If the singularity C n /G admits a crepant resolution, then there is a junior element g ∈ G.In fact, [18, Par.4.5] conjectures that under the same hypotheses, if the singularity C n /G admits a crepant resolution, then any maximal cyclic subgroup of G contains a junior element. A counterexample to this conjecture is however presented in Remark 8.15. In this section, we prove a weak version of that conjecture. We phrase it in an analytic set-up, as our later applications call for that, but the proof works in the affine set-up just as well.Proposition 2.7. Let G ⊂ GL n (C) be a finite group acting freely in codimension 1 on C n , and let U ⊂ C n be a G-stable simply-connected analytic neighborhood of 0 ∈ C n . If the singularity U/G admits a crepant resolution X G , then the group G is generated by junior elements.Note that a singularity admitting a crepant resolution is Gorenstein. By [20][42], the existence of a crepant resolution X G thus implies that G ⊂ SL n (C).In order to prove the proposition, we need some background in valuation theory.