I would put this in much stronger terms. The theory unambiguously indicates that changing from block reward to txn fees will fail under the current design unless mining becomes monopolized.
We will return to this topic in 30 years, when it will be possible to verify these claims.
For now, stick to the matter at hand. And the matter is, we don't need any fix for a problem that neither does exist yet or it is certain it will ever exist.
My guess is that a single entity will acquire 51% sometime during the next block reward cycle. (that is during the period when the block reward is equal to 25, starting in December of 2012). If the individual is malicious (admittedly unlikely), the developers will likely be taken by surprise and won't have time to save the blockchain.
51% attack won't destroy the blockchain...
Also 51% is actually not enough to perform massive scale attack. You should have few percent more, which is highly unlikely (for example, if deepbit became the adversary, large percent of users would stop using it instantenously).
And we already have a contermeasure against 51% attacks - decentralized mining pools. If only people used them more often.
So increasing inflation by any means is not necessary. What kind of solution is that ? This is
common thievery, nothing else.