The law of entropy is, in truth, nothing but statistics - in any given quantum-level system event, the law of entropy can be broken, it's just -very- slightly less likely than the alternative of being conformed to.
Now the fascinating thing is how this interacts with the two principles entropy can be conceptually derived from.
Functionally, the entropy of a system is a measurement of the energy which cannot be used.
Conceptually, there may be more meaning.
The first is the simpler of the explanations, and says, basically, that energy has levels (potential work), and that energy is constantly decreasing on this scale - effectively, that the amount of work that can be extracted from energy is finite. (Intriguingly, there's both a minimum and a maximum rate at which this can occur.) This explanation is more or less hand-waving the issue of entropy away by describing it.
The second, more interesting, explanation, relates to informational theory. Without getting too deeply into the details, informational theory - as relates to physics - holds that information cannot be destroyed; information being a description of the states of matter that led to the current situation. (That is, the end state of matter must describe every state of matter leading to it, systematically.) This means the number of potential states is in constant decrease as time goes on - and energy is constantly being bound up in a descriptive process.
Modern physics is very firm about the conservation of information; the only theoretical exception existed, according to Stephen Hawking, around singularities, and he has withdrawn this exception.
Just one problem. Informational Theory as it pertains to physics is an absurdly good example of a patently unscientific theory. It is literally impossible to test; how do you demonstrate that the history of an object was not accurately recorded within the universe? You have to accurately record it in order to do so, and, well, you're doing so from this universe. The only interesting description of entropy depends upon a "theory" which cannot be tested.
Inconvenient, to say the least.
More inconvenient that quantum theory suggests the law of entropy -can- be broken on minute scales, because this would break information theory, if it held true. Indeed, quantum theory and information theory clash very fundamentally in one very fundamental way: Information theory holds that the amount of information is constantly increasing, meaning physical constraints on the universe are constantly increasing. At some point, the amount of information is going to violate the uncertainty principle; that is, there is a point at which no more information can be constructed without "using up" uncertainty.
It is, of course, very messy. Personally, I suspect the law of entropy is more of a suggestion.