Is there a balancing point to the difficulty?
You know, at one point I would have just given a flat "Yes" as a reply. Now I'm not so sure.
The two primary limiting factors on difficulty are: the (recurring) cost to mine, and whether mining at a given difficulty can recover that cost. Presuming that the cost of bitcoin stabilizes at some point in the future, and that ASICs are the endgame as far as efficient hashing equipment, then there should be some difficulty such that the miners across the planet, collectively, barely break even on the cost while mining with the latest ASIC technology. And that would seem to be the natural equilibrium point, right?
Except that processing power still seems to be following Moore's law, and AFAIK there's no limit in sight as to how efficient ASICs can get. As ASICs continue to get more efficient, difficulty will continue to increase.
Plus, the release of cheap, plug-and-play USB miners changes things too. When people can pay a reasonably sum to mine, many will, for a variety of reasons, be willing to do so at a small loss. With such devices in existence, this means that network difficulty will be slightly higher that what it otherwise would be due to extra, non-profitable mining that continues supporting the network.
And then finally, should bitcoin become widespread enough, you'll have companies who mine, not for profit, but just so they have ready access to transaction processing. For example, Wal-Mart could conceivably setup a mining center so that it could process millions of transactions a day without paying anyone else processing fees, particularly if the cost of the mining center is less. Other large companies could follow. Collectively, this would add even more non-directly-profitable power to the network.
The upshot of all that is: while the rate of increase in the difficulty will eventually slow (probably,) I don't see the difficulty going down anytime soon unless bitcoin were to become obsolete, or the price were to crash and stay low.
But that's just it... as technology continues to improve such that reasonably small mining devices can continue to be produced, that will never be the case. Because even if it's not profitable, you will always have small miners adding to the network hash rate for reasons besides pure profit.