So people would create a routine circuit where they have dollars, buy bitcoin on US exchange. Send bitcoin to foreign exchange, sell bitcoin for that national currency (sometimes dollars). If possible, that exchange would allow wiring out, back to the user's real bank account. In other cases that exchange would have a bitcoin or national currency market of another cryptocurrency, like litecoin, where an arbitrage opportunity MIGHT still persist.
The litecoin would be moved and liquidated back on a US exchange.
Got to calculate it yourself and figure out why the arbitrage opportunity is still there. Usually there's a good reason. Sometimes there's a good reason that you are exempt from, and in those moments you borrow as much as a can get your hands on and make 5% profits as much as possible.
The difficult part is in transferring non-bitcoin currency to and from the exchanges.
If you for example buy bitcoin in the US and sell in China, then you end up with a surplus of CNY on a Chinese exchange's account. Getting this converted and transferred back to the US will incur exchange and transfer fees. The transfers might also take weeks - which will limit you to cycling small amounts of money slowly, or risking a lot of money at once if the exchange "has banking issues".
In the end, even a 5% or 10% price difference might not be worth running the arbitrage cycle on difficult to deal with exchanges in strange countries.
My current best option is hoping that a) the Winklevoss ETF gets approved and b) an option chain develops for it, in which case I will buy puts consistent with my belief that the long-term FMV for the ETF's assets is zero. (Exchange-traded options in the US are guaranteed by the Options Clearinghouse Corporation, which is why -- absent global calamity -- you can be very certain that e.g. puts on Google are worth money even if Google goes out of business or the people who sold you the puts are insolvent when their broker hands them an exercise notice.)
There is theoretically a DRW subsidiary which does over-the-counter options but they have a $25k minimum trade size. It would also require a lot of due diligence for me with regards to counterparty risk; it's not clear whether their options would reliably compel performance or that they would be in a position to backstop the failure of their clients in the event of total systemic collapse of the Bitcoin markets. It's also not clear to what extent that DRW would backstop them if they were insufficiently capitalized. (For an illustration of why this matters, read The Big Short, for the amount of heartburn that various folks betting against the housing market went through when it became obvious that they were entirely right on the merits but that some of their counterparties were almost certainly going to go bankrupt due to how wrong they had been about the housing market.)
(I've spent way too much time thinking about this, since "Bitcoin will eventually fail catastrophically" is the biggest answer to "A belief about the future state of the world which I hold strongly after reflection and which doesn't match the beliefs of some of the smartest people I know", which sets off my "Either I'm wrong or I should be betting against them" antennae.)
The bitcoin network decided to strip the VM of features that made it valuable in the future (eg, ability to make transaction claiming represent the solution to a computation) in favor of safety and simplicity now. I don't believe the behavior of the network in resolving blocksize issues indicates the kind of political will needed to fix that.
So my bet is that the next iteration (or two or three) on the ideas behind Etherium will eat their lunch. Because having tokens to reward contract execution or a computation has a fundamental longterm value (and is similar to how real currencies function, in that executing a contract or doing a computation earns you the ability to have that done for you). I don't believe that just securing the ledger has long term potential.
both uses can be crushed very easily and then you will end up with a highly devalued good.
Firstly, the reality is that other people (lenders), rather than exchanges, are lending out their money, for a certain amount of interest. Exchanges have nothing to lose. Quite the opposite: they earn regardless of whether longs and shorts are successful or not. Take Poloniex as example: they collect a 15% premium on interest earnings. This is of course next to the premium they collect in the form of regular trading fees.
In nearly all cases this means that lenders don't lose money. Borrowers do because they pay interest. This is why you can only borrow after you allocate collateral, which is used to pay for said interest, and to collect profits, cover losses, and for forced liquidation in the case that the borrower's trades are about to lose more value than is covered by their collateral. This applies to shorts and longs.
The only chance this becomes problematic for an exchange, I think, is when a forced liquidation does not cover all of the losses and some debt stays open. In this case either the lender does not receive the right amount of interest, or, what makes more sense to me, is that the exchange will cover the debt and try to claim it back from the liquidated borrower.
Secondly there are a good number of exchanges that do have very decent track records and aren't as scammy as you claim. There were a few scammers out there, I'll grant you that - companies like Cryptsy and of course MtGox just stole incredible sums of money. And yes, this can happen again with other exchanges. Be that as it may, this still does not have anything to do with the possibility of shorting in particular.
What you say of the exchange business model not being exposed to price movements in principle might be true of properly capitalised and regulated exchanges, but its less evidently true of the exchanges that actually exist.
The most recommended place to short Bitcoin is/was Bitfinex...
This sounds more like speculation than arbitrate. With arbitrage, you transfer liquidity from one exchange to another by having a short leg on the higher-price exchange and a long leg on the lower price one. This is great for Bitcoin because it evens out the price differences between the exchanges, and at the same time transfers the liquidity of one exchange to another one, while the arbitrageur makes a profit.
I don't see how leveraged trading can replace this. Arbitrate is, essentially, taking orders from one exchange and selling them on another one, thereby matching a buyer and a seller on separate exchanges. It's a genuine service to the market.
It ins't necessary extreme speed, in fact for me it took more than a month to settle.
I would first wait my program notify me there was a 20% or more profit opportunity (due to the slowness and risk).
Then I would buy btc with credit card on some usd exchange (usually mtgox, not always), then I would sell btc for brl on a Brazillian exchange.
I would pay the credit card with my brl when the bill came... this could take more than a month, and I would pay lots of taxes, but I would still turn a profit.
The reason I stopped doing arbitrage was that to me was too time consuming, because every month the payment methods would change (for example someone one month accepted paypal, then only cc directly, then only paypal again, then some paypal competitor where I had to register...)
I have been doing bitcoin arbitrage for a while with https://github.com/butor/blackbird/ . The project seems to have gotten some more attention recently for what ever reason.