Currently a botnet could generate 99.9% of bitcoins per day, but if bitcoin generation payout was variable, they could generate over 9000% of how many bitcoins are generated per day without them.
9000% of the now-cheaper-coin because so much of it is available after they joined. It all balances out in the end, and in both cases they have an incentive to join the network. Actually, under the current model they have even more incentive because the coin can never get cheaper!Botnet operators value their CPU cycles, so being able to generate over 9000% until they have as many as they currently want and then being able to stop and use their CPU cycles for other purposes is much more advantageous than having to use their spare CPU cycles to only generate a capped amount per day.
To be honest, I don't see what more harm a botnet can make if we changed to the suggested model. Economically, it would be just like 1,000 honest members joined the network. This can even help the project with some free advertising on security blogs 
The number of blocks generated per day can not be easily increased or decreased because there has to be enough time to propagate the blocks to all other nodes while still being frequent enough to allow timely transactions. What can vary is the amount of bitcoins per block. The amount of bitcoins awarded per block can either increase, decrease or stay constant over time. If every person had an equal amount of CPU cycles at his disposal, I would prefer for the amount of bitcoins awarded per generated block to increase proportionally to the amount of CPU used to generate bitcoins. As it is, I'm in favor of the amount staying constant over time.
How about a middle solution then? For example, the default difficulty be adjusted to about 1 block/CPU/day (assuming an average 2.4 GHz processor), but in case more than X blocks/day was being created (X = the maximum # of daily blocks afterwhich there is no enough time for propagation across the network), then difficulty gets automatically adjusted. Practically, it would end up being like my second-best and your best model. In other words, very close to the current situation except that we assume the current 4-years-period is extended to forever. So let's at least agree that the removal of the 4-year thing and the 21M limit is essential to avoid an ever-deflating currency.He was referring to the price of goods decreasing as value of a bitcoin increases. Eventually the value of bitcoins will go high enough and the price of the goods will go low enough that the person will feel rich enough to spend his bitcoins. To take it to an extreme, if a person has ten thousand bitcoins and can buy the TV of his dreams for one bitcoin, the value of having the TV now is greater than the future potential increased value of the one 10,000th of his total bitcoins.
You only switched the problem to another user then: The TV seller! He would also keep this coin until he has 10,000 coins. The point is that it's very hard (albeit not impossible) to convince people to spend or transfer a coin they are pretty sure will be 19% more valuable within a year. Moreover, economic-savvy folks might as well choose not to join the network if they predict this will happen, or join it with the intent of "hitting-and-running" a quick profit before the system's failure becomes imminent, and they will be right.