Operators - how to motivate they must be linear ? Is this comment a hint? Is there a way to motivate, retrospectively, that observables must be representable by 


*

*linear operators 

*on a Hilbert space? 


Specifically, there seems to be a hint to something in the accepted answer to this question. There, the author writes that 

"[...] and since these commutators satisfy the Jacobi identity, they can be
  represented by linear operators on a Hilbert space."



*

*Is this true? If an observable can be written as a commutator (like the three coordinates of angular momentum), does it automatically follow that it corresponds to some linear operator? If yes, how / is this a theorem with a name?


Thanks for further hints, and for all hints so far. Quantum mechanics is still really strange to me, additional motivation for one of the postulates would be edit: is nice. 
If you want me to clarify further, please say so.

This question: "How did the operators come about" seems related, but the answer doesn't help since it starts from the postulates. My question is about possible motivations one could give for one of the postulates.

Revision history part 1: The question was quite general before, it also asked "where did the operators come from" which aready has a very good answer here.

For completeness / Revision history part 2:
Another approach might be via Wigner's theorem, that any symmetry operator on a Hilbert space is either linear unitary or antilinear antiunitary (and then one could maybe go from symmetry operators to their generators, and find out that they correspond to observables). Originally the question was also whether this line of argumentation works as well. But probably, this is covered in the historical sources, given here and in the comments.
 A: From the physical standpoint, it is natural to suppose that it is possible to manipulate observables mathematically in a suitable way. It should be possible to sum them, multiply them, and scale them in order to obtain new observables. In addition, it is often convenient to extend the concept of observable to complex objects for which it is also possible to do "complex conjugation" (abstractly, called involution).
Complex numbers, or complex valued functions, allow the above manipulations, and could be taken as observables (and the latter are in classical theories). To describe quantum systems, however, such observables should also be non-commutative, since effects due to non-commutativity are observed experimentally.
Mathematically, objects with the properties above form an involutive algebra, abelian if they commute, non-abelian if not. There is a theorem, of Gelfand and Naimark, based on a construction of Gelfand, Naimark, and Segal, that says the following.

Every *-algebra is isomorphic to an algebra of linear operators acting on a Hilbert space

It is therefore natural to represent quantum observables as linear operators. For classical (abelian) observables, the theorem shows that the operators in that case are multiplication operators, and that abelian algebras could also be represented as an algebra of complex-valued functions acting on a suitable topological space. The latter representation is the one usually chosen for classical theories.
