display | more...
With regard to the second "possibility":
It will never end. Ever. This has some amusing ramifications, including eternal recurrence. This would be the case if the universe were to somehow last for an inifinte amount of time. As demonstrated by Henri Poincare, if the particles in the universe are given an infinite amount of time, they can take up an inifnite number of configurations, and therefore entropy will sometimes, in the very far future, be decreased.
This is not what Poincaré demonstrated. Poincaré showed that if your phase space either is compact or has finite volume, then paths are strongly recurrent. This indeed would mean that behaviour would be almost cyclic.

But for the universe "never to end ever" in the sense of simultaneously avoiding a big crunch and heat death, it must have an infinite amount of negentropy to play around with. In particular, you can forget about finite volume conditions -- this would require us to observe finite volumes with infinite negentropy in them, which we certainly do not. And compactness conditions are much the same, for any remotely reasonable geometry.

Infinite negentropy requires an infinite universe, which is not absurd. But then Poincaré's results don't apply.