When the BFL chips from this last batch all go online (although anybody's guess when that will be, they've delayed shipments so many times) the network difficulty will be so high that only ASICs will be profitable.
But look at mining this way: you have to leave it plugged in and running all day every day for it to make money. People are only going to do that (and pay for the electricity) when it's worth it to them.
Say a mining rig costs $X. if it can produce $X in bitcoins in only a few months, then EVERYONE is going to want to spend $X to get one. It's a no brainer. But if the difficulty is so high that it takes one or more years to produce $X worth of BTC, more people will drop off. And people that do have a rig might not run thiers all the time, since it''s not very profitable.
So if you think the initial gold rush is over (which I argue it is, since ASICs aren't a secret anymore), then with human greed expect that whatever you spend on a mining rig it will take 12-18 months to recoup your investment. After that the profitability will go down because of the increasing hash rate.
If it's less than 12 months for mining rigs to pay for themselves, more people will buy them. If it's more than 18 months, then less people will buy them.
So eloquent how you described that +1 - FWIW I diversified my approach to btc and cryptocurrency - I have 2 X6500 FPGAs mining BTC and sometimes FRC, I bought 2 1 year shares of cloudhashing 3 5gh singles and a 50 gh single for starters. I've seen varied gains from using the 6500s but they are at least light years past the single 5450 I had been using prior.