If trying to make a practical judgment (i.e. should I use a lighter or heavier wheel for this ride), you have to assess the amount gained or loss by the flywheel effect in relation to other gains and losses from added or subtracted weight.
I suspect that flywheel gains and losses would be quite small in comparison to air, rolling, and slope resistances. Hence, in climbing the added weight would slow you down (although we're talking 200 grams out of say 80kgs (bike plus rider). That might make the climb one second slower over a 5 km 5% grade climb. The amount regained in the descent would be even less, because descending takes much less time to accomplish than climbing. Any gains or losses from the flywheel would be a mere fraction of that
Furthermore, as already mentioned, any energy preserved by the flywheel had to be put there by the cyclist in the first place, which means that it would come out equal, minus efficiency losses (3-5% in a well-lubed drivetrain), both as the energy goes in and as it comes back out.
I think the theory is bogus for the later reason alone. Use of heavier wheels has surely to do with their aerodynamic qualities and not with their added weight for their flywheel effect.
Cheers,
Jamie
P.S. Vlad, no matter how I read it, your comment about the small-mindedness comes across as pretty harsh. BikeForums being "known" for having smallminded participants insults the group no matter how you read it. What good does such a comment serve? Did you think that you'd actually provoke a better discussion that way? Hmmmm....