The concept of entropy is only an approximation: it can only be defined exactly in the thermodynamic limit of an infinite number of degrees of freedom. The same thing is true
of all thermodynamic concepts, including phase of matter, equilibrium, etc. Such an approximation is only valid within certain limits...when pressed outside of those limits it yields paradoxes, nonsense, etc., as first pointed out by Sir James Jeans when he talked about the paradox of the thermodynamic law of increasing entropy contradicting the time-invariance of mechanics. (It must be remembered in this context that Boltzmann's programme was precisedly to deduce the laws of thermodynamics, including this one, from the laws of mechanics at the micro-level.)
It follows from this that the so-called arrow of time is merely a useful approximation which is good for some practical purposes, but not necessarily all purposes. Therefore,
The answer is Yes: it is analogous to quantum measurement since the same thing could be said about quantum measurement. See my
for more details and references, also http://arxiv.org/abs/1108.3151 for Jeans's approach to these thermodynamic questions.
Also relevant to « the arrow of time,» are Feynman's thoughts on quantum measurement:
Feynman, in his Path Integrals and Quantum Mechanics, said that the reason why Quantum Mechanics is deterministic when you look backwards, to the past, but only probabilistic when you look forward to measurement results in the future was «undoubtedly » due to the fact that measurements rely on amplification, and that what was lacking so far was a statistical mechanics of amplifying apparatus to understand this. Cf http://arxiv.org/abs/quant-ph/0502044 for exact references and more discussion of this. See also Allahverdyan, Balian, and Nieuwenhuizen, http://arxiv.org/abs/cond-mat/0203460
In my opinion, the very sane view of Jeans's about the status of thermodynamic concepts has been lost sight of (even by Balian) due to the lack of progress in proving Khintchine's conjectures.