Sure, but the reward sizes are set in stone in the protocol and follow a predetermined path. The maximum block size will have to adjust to balance security and transaction demand.

There is a small cost to adding a new transaction. You have to store the transaction until it is spent, validate the transaction, and recompute part of the Merkle tree for the block. There will also be big network, disk, and electricity costs for generating blocks.

Block size limitations indicate that the network is unanimously willing and able to download, upload, and store blocks of that size every 10 minutes. It will probably stay at the current 1MB until Bitcoin has a solid backbone of generators. Then it can increase rapidly.

Some generators will set fees to the minimum, and some will set fees to the highest profitable level in the hope of raising it further. Most will be set somewhere in the middle. The generators will be saying, "We want to be paid on average this much," and the users will be saying, "We will pay only this much on average." This creates a really nice effect: your transaction will almost always clear eventually, but more fee = more speed.

The "minimum fee" for a generator will usually be the fee that causes the most fee-transactions to be included in the block. Volunteers or people who feel that the generation reward is enough might have a minimum fee of 0. This will happen less and less in the future, though, because costs will rise, inherent rewards will be reduced, and you'll therefore want to create blocks with at least some non-free transactions.

If the network overcharges, competitors will come in that claim the leftover transactions. If the network undercharges (a high-fee transaction exists, but it is not included in a block in favor of a cheaper one), competitors will come in with proper fee-based prioritization.

When the coins generated per block is low, users may not be willing to pay enough for generators to ever be profitable. Transactions will be really slow in this case, and the users will hopefully change their minds. If they don't, the least efficient generators will leave, the difficulty will be reduced, and generation will be profitable again. The users will be saying, "I'm not paying for that much CPU power! Reduce it."

It's a backward-incompatible change. Everyone needs to change at once or we'll have network fragmentation.

Probably the increase will work like this: after it is determined with great certainty that the network actually can handle bigger blocks, Satoshi will set the larger size limit to take effect at some block number. If an overwhelming number of people accept this change, the generators will also have to change if they want their coins to remain valuable.

It might also work in reverse, where almost all generators decide to reduce (or raise) the max block size. If clients want to have any real protection from double-spending, they also need to switch.

The limit needs to increase at some point if Bitcoin is to become mainstream. 1MB is not an awful lot of transactions, so not increasing it would raise fees to unreasonable levels.

Most transaction are free, but you can make it go faster by paying transaction fees.

Something I realized that is slightly related is that in the future the time between blocks will become more regular. Currently the reward is the same no matter how long it has been since the last block. When fees make up most of the reward they will increase linearly and as the pot builds more and more generators will come online. Of course some will not bother turning off after a block is found, but those with high electricity costs or with something else to compute will switch on and off.

Hmm, I can't think my way out of this problem.

Blocks will always be solved. Sometimes way too easily.

There are two cases. One is when all transactions fit in the blocks. Then there is a tragedy of the commons. Transaction fees drop towards zero and there is hardly any incentive to generate. This is the case now. Transaction fees are zero and the only reason people generate is coin creation.

The other case is when all transactions don't fit in the blocks. Then transactions will compete to get into the blocks using transaction fees. The maximum block size must be continuously adjusted to keep the transaction prices stable. The only way to change the maximum block size is through a lengthy political process of debate, decree, network fragmentation and majority agreement.

Correct me if I'm wrong here, but it is right now a very rare situation where blocks are even close to the maximum block size?  In fact, on average most blocks don't even have transactions included in them at all.  In other words, all they are doing at the moment is purely generating coins.

If there is a situation where there is a huge pile of transactions backing up, perhaps the ones with fees could get move to the front of the pack, so to say, but the rest would generally be cleared in subsequent blocks.

There certainly isn't anything stopping you from modifying your source code on Bitcoins and only accepting transactions with fees for at least blocks you are able to mine and create.  Essentially, if you are going through the effort to generate a block, you ought to be paid for any transactions in that block, at least if you think that way.  If enough people don't want to process transactions without getting paid, those without a fee simply won't get processed.  It really is that simple.

For myself, I'd love to see this as a option in the User Interface itself where you could set the rules up in some way for how you would personally want to process transactions, and make it more of a market-based approach.  There is room for multiple rules regarding how transactions are processed.

No, there will be instititutions with a high value to protect.  Costs of protection are a 'tax' of sorts, but they are unavoidable.  They are not a tragedy of the commons scenario, each is still looking out for his own interests, and his interests benefit others.  It's a postitive externality.