But why is that a special case? The way I understand it: the normal series can always be made such that all composition factors are simple, but then the composition factors are both simple and Abelian, and hence (isomorphic to) [itex]\mathbb Z_p[/itex], i.e. the composition index is p (= prime)...

It is only for a finite group that you are guaranteed to have a composition series. For an infinite group, there may be no normal series where the subgroups are maximal. For instance, ##\mathbb{Z}## cannot have a composition series, since it is not itself simple (every subgroup of ##\mathbb{Z}## is itself isomorphic to ##\mathbb{Z}##).