I have to admit that this is a concern that never occurred to me. But it seems implausible in the extreme. There are a hundred billion XRP, or a hundred quadrillion drops. Let's say 1/10th of a penny per transaction is small enough that nobody need be concerned about it being a limiting factor. For one drop to be worth 1/10th of a penny, the total value of all XRP would have to be one trillion dollars.
Your example is flawed. You say that the transaction fee will be only 1/10 of a penny. What happens if Ripple grows and new XRPs are not handed out in the system? The value of XRPs will go up without bound. Especially because, once an account is funded the XRPs are "locked in." From an analysis perspective they are effectively destroyed since they can't leave the account. That means that the rate of deflation is many orders of magnitude greater than the rate of XRP destruction due to transaction fees. "Implausible in the extreme" is only true if
no one controls the bulk of the XRP supply.
If that ever did happen though, the divisibility of XRP could be changed by consensus. Alternatively, we could use a scheme where you prepay for a number of transactions and then can perform transactions without a fee until the account's "transaction credits" drop to zero.
I keep hearing about how easy it is to change the transaction fee through the consensus mechanism. But no one has explained exactly what criteria a gateway will use to determine if the fee needs to be adjusted. How do you implement these functions:
bool shouldTransactionFeesDecrease (NetworkState s);
For that matter, how do we implement this function:
bool isTransactionSpammy (Transaction t);
But I think this is a discussion best left in another thread. The bigger issue is the XRP distribution.