The expansion of the universe
It has been known since the late 1920s that light received from very distant galaxies is red-shifted from its expected wavelengths. This has generally been interpreted as an expansion of the universe. Since the red shift increases the more distant the galaxy is, the conclusion is that there was a period in the past when the universe was highly compressed compared to its condition now. Therefore, the universe appears to be running on irreversible lines.
However, some theorists have speculated that the gravitational attraction of matter will eventually reverse the expansion, resulting in a collapse known as a Big Crunch. Einstein considered a series of expansions and contractions, but a potential problem with that was pointed out by Tolman, who argued that whilst the laws of mechanics might be time reversible, the laws of thermodynamics are not, and they say that the entropy of the universe can only increase.
Our view is that thermodynamics has to be discussed contextually. All the evidence points to that. Statistical mechanics requires a definition of microcanonical ensemble, coarse graining, and such like concepts that depend on choices made by observers. In the same way that quantum cosmology can be criticized for being too loose with the concept of ensemble and observer [Fink & Leschke, 2000], it seems to us to be equally inadmissible to discuss the thermodynamics of the universe in a contextually incomplete way.