If the network hashrate were steady at say, 1 quadrillion hashes per second*, the difficulty would adjust so that exactly 2 weeks pass per 2016 blocks. If the hashrate then grows by 10%, blocks will start coming (on average) every 9 minutes, and it will only take 12.6 days to reach 2016 blocks, so the difficulty will go up, by about 10%.
Right now, the network is always growing, so the difficulty is always adjusting up to balance that out. The more power we add, the higher the difficulty gets. It is a feedback system fighting to keep the block rate as close to 1 per 10 minutes as possible.
New coin generation comes from blocks. Because we can't fool the difficulty, the blocks come out on schedule, and thus the new coins come into existence on schedule. Well, we can fool the difficulty a little. It doesn't overshoot us, so if we grow really fast on the right day of the cycle, we can bang out a bunch of blocks really quickly before it catches up, which brings future blocks forward in time, more or less permanently, by a bit.
* We are really closer to 5 or 6 quadrillion hashes per second, right now.
Right now, the network is always growing, so the difficulty is always adjusting up to balance that out. The more power we add, the higher the difficulty gets. It is a feedback system fighting to keep the block rate as close to 1 per 10 minutes as possible.
New coin generation comes from blocks. Because we can't fool the difficulty, the blocks come out on schedule, and thus the new coins come into existence on schedule. Well, we can fool the difficulty a little. It doesn't overshoot us, so if we grow really fast on the right day of the cycle, we can bang out a bunch of blocks really quickly before it catches up, which brings future blocks forward in time, more or less permanently, by a bit.
* We are really closer to 5 or 6 quadrillion hashes per second, right now.
And to put into context how effective that is, we've already experienced it at a time when there mas maximum incentive to exploit this. Last fall, right before the block reward halved to 25, with the first ASICs. I haven't looked recently but at the time there was panic that there'd be a 10-day period where blocks were coming every couple hours at 50BTC each. This didn't happen despite maximum reward opportunity for collusion. And likely there WAS collusion, but it wasn't feasible to do more than they did. IIRC, faster than about 7 minutes/block has never been the average for a given set of 2016.
Economically this is also of only worth spending time on, if you mine for a living. For users, the effect on supply and demand for BTC is completely independent of hashrate, and a 2016 block period of overproduction cannot create more than 50,400 BTC, regardless of how fast or slow. That many could be dumped into circulation in an instant with only a blip of effect on exchange rate, and IRL a mining effort would want to maximize profits off that set, so they'd never sell all at once.
If the trendline was for BTC's 3-month appreciation vs. USD was 12%, regardless of instantaneous effect (within the duration of the selloff), that would only depress the RoA by a max of 0.42%, ignoring some short-term psychological effects if that instantaneous window is small enough for the sale to be anomalous. If no news, community assumes an old hand retired or died leaving his BTC to heirs who claimed cash. If accompanied by bad news there may be amplification f the suppression, but not much within a 3-month interval, and less when even longer periods are used.
Interesting...so what does the nominal network hashing power need to be then?
Just as the above hashrate is the maximum difficulty the current mechanism can adjust to (and we'll have switched to another one already should advances ever take us near that, which certainly seems unlikely to happen within many, many lifetimes), the minimum difficulty that is permitted is one in which the probability of winning is 1 guess out of 23,282,709,094 (rounded to a whole number in decimal after conversion from the 12-character floating point hex value).
Thus, the minimum global hashrate to maintain a 10-minute block discovery average is approximately a guess every 0. 0000 5195 2717 seconds, which is 19,248.2714071 hashes/second. Factor for maybe a 1% orphaned block probability to be safe and you get something around, lets say 19.5 khash/sec for a pretty human-readable minimum. Again, if we ever approach this, something very unexpected has happened, in this case presumably the cost of acquiring electricity went up by an unimaginable amount in a few days, and we'll have switched to something else as we try to get a grip on such a sudden change of events in the course of human history.
Just as the current system of value exchange and transmittance is inefficient in light if current technology, thus bitcoin, an uncorrectable loss of widespread and cheap Internet-like systems would create a similar situation for bitcoin, but it would be possible to adjust the system for post-apocalypse operation in the interim. A few changes to the protocol and humans could manually operate the system. We'd be quickly trying to split the blockchain into geographic regions, and would ultimately wind up with a system not completely unlike what we had before bitcoin.
