And now imagine Satoshi-dice translated in 100 languages and used daily by 100 million people. And now imagine imagine another 100 Satoshi-dice type of service. That's going to be a problem, no matter how big the block size is.

We're all keen to see efficient protocols built on top of Bitcoin for things like micropayment channels (which allow lots of fast repetitive satoshi-sized payments without impacting the block chain), or trusted computing (which allows offline transactions to be carried around in long chains until final resolution). Also the payment protocol should eliminate the most absurd abuses of micropayments like SDs messaging system. These things fall into the class of "no brainers" and were discussed for a long time already.

Other more exotic ideas like Ripple-style networks of payment routers using contracts don't seem against the spirit of Bitcoin if they keep the low trust aspects of the system.

At the same time, as evidenced by the disagreement on this thread, there are too many unknown variables for us to figure out what will happen ahead of time. The only way to really find out is to try it and see what happens. If Bitcoin does fail to scale then the end result will be a smaller number of full nodes but lots of people using the system - this is still better than Bitcoin being deliberately crippled so it never gets popular because even if the number of full nodes collapses down to less than 1000, unknown future advances in technology might make it cheap enough for everyone to run a full node again. In the absence of a hard-coded limit the number of full nodes can flex up and down as supply and demand change. But with a hard-coded limit Bitcoin will fail to achieve popularity amongst ordinary people and will eventually be forgotten.