But what if the apparent probability after n trials does not converge as n grows arbitrarily large?

If we observed such a system in nature, what would its "probability" mean?