Sadly Bitcoin is fundamentally unstable. If the difficultly ever stabilised it is vulnerable to a mining efficiency attack that de-stabilises it. It's one reason that I'll be staying clear even if it does become more mainstream.
Or is there something I'm missing?
It's one of the quirks of capitalism that markets remain efficient only to the extent that people believe that they're inefficient, and liquid only to the extent that people are misinformed about an asset's value. To see why, imagine a perfectly efficient asset market with perfect information, where the goal of every participant is to maximize their financial returns.
In such a market, all participants know everything that all other participants do, they know that all other participants know everything they do, and they can transact instantaneously without cost. Prices in this market converge instantly upon the "true" value, which can be agreed upon by all market partipants because of their omniscience.
Such a market has zero liquidity. Why would anyone transact if they know that a security already has the correct value? For there to be profit potential, the person who sells to you must be misinformed about the security's true value, which contradicts the assumption of omniscience.
The interesting thing is that the more market participants believe in the market's efficiency, the more inefficient it becomes. With little incentive to do their own research (why bother, when the current price reflects more information than they could possibly amass?), the underlying fundamental value of the asset can drift further from the consensus price without anyone noticing. This increases the profit potential for people who assume that they know better than everyone else. Eventually, a few of them get rich, the consensus breaks down, lots of people lose their shirts, and the market becomes efficient again.
The result is the business cycle. Periods where people believe in the market's efficiency are bubbles (when the consensus price is above the true value) and panics (when the consensus price is below the true value). Periods when they don't are crises, where the prevailing momentum swings but this change in opinion is unevenly distributed.
Lots of things fundamental to the modern world are unstable. The whole concept of a financial asset is one of them. I can think of a number of individual industries that have an unstable dynamic, but I'm not going to name them because I believe there's profit potential in them. Public opinion is yet another one.
If difficultly is stable then what matters is how efficiently you can mine, i.e. how many bitcoins you can mine relative to how much electricity you spend.
Well if you control say a third of the pool, then you can obviously capture a third of the bitcoins and spend a third of the electricity.
Or, you can mine alternate difficultly periods.
When you mine, the next period the difficulty will be 1. Then you chose not to mine and the pool loses 1/3 of the hashing power, so next round the difficulty will be 2/3. Then it adjusts back and it will be 1 again.
The result is that you get to mine at a lower difficulty than the rest of the population. The difficulty increases from your hashing is paid for by everyone else and not by you, then you get the enjoy the reduced difficulty from others not mining as fast.
This becomes de-stabilising because the rational thing when faced with this is to join you on only mining the easier periods which further makes the problem worse.
This effect starts happening with any amount of hashing power, even with 1% hashing rate you would see a small efficiency improvement by mining alternate difficulty periods.
Plenty of systems work fine like this in nature - it's pretty much a textbook predator/prey dynamic, and yet carnivorous life has been going on for several hundred million years on Earth. It would be a problem if the dynamic was for the number of miners to drop to zero or explode to infinity instead, but this is a negative-feedback cycle, not a positive-feedback one.
You are wrong as characterising this as negative-feedback, this is positive feedback because if you are presented with a situation where in even "weeks" (which now last 6 days) the difficulty is 10% less than odd weeks (which now last 8 days), then you will join mining only the easy weeks which further feedsback to more distortion.
edit to add: I realise the adjustment period is not aimed to be 7 days, this is for easier illustration.
The question is what is the "activation energy" required to initiate the oscillations. Is a large miner (how large?) likely to attempt this knowing the ultimate outcome?
And more importantly, can it be fixed? The protocol isn't set in stone.
This attack can be pulled off in economically satisfying way only then your electricity costs dominate your overall costs, i.e. you do not pay rent, wages etc. And this is false.
Given miner's situation (he pays wages and rent) to use this attack effectively miner needs to represent significant part of hashing power of network. This means that his stakes are high and engaging in destructive behavior may undermine reputation of Bitcoin as network, BTC price and ultimately miner's own business.
In case lower scale attack (manipulation of hashing power of a pool) pool and his operators are there to punish such behavior.
EDIT: language
What happens is that in higher difficulty periods your competitors who keep running constantly are making 1800 coins each. When the hash rate drops and you come on everyone gets 1200 coins each. Your expected revenue becomes 600 coins/period. The result of your strategy relative to just running constantly is halving revenue.
The only variance introduced is the speed at which new blocks are generated in each period.
The important point is that an attacker could use this just to wreck the cryptocurrency in question, although they may simply buy more coins at the lower price and profit from any rebound.
You can't rapidly jump between difficulties like that.
It's about the fact that you mine more efficiently not suddenly mining a lot more bitcoin. (You actually mine fewer blocks overall but at a larger reduction in running time costs than the reduction in mining.)
Mining "more bitcoin" or mining "more efficiently" don't matter. What matters is the speed at which 2016 blocks are produced network-wide.
Everyone mines at the same difficulty. If you choose not to mine for a period (and the difficulty adjusts to a lower level), you're just encouraging more miners to fill your void - your absence represents new opportunity for miners who might not have been profitable at the higher/original difficulty.
Even if new mining doesn't increase difficulty to the original level, you're always mining at the same difficulty as the rest of the population, so your supposed gain is everyone else's gain, meaning it's effectively non-existent.
In other words, just because you can affect difficulty by temporarily choosing not to mine, your absence (and lower difficulty) does not decrease participation of other miners; if anything, it will encourage more mining and ultimately cause the difficulty to increase for everyone.
Do you mean alternate difficulty periods, as opposed to alternate blocks (whch would mean switching on/off every 10 minutes and would not affect the difficulty significantly.