I meant to answer this last time you asked, but I didn't get to it because, to be honest, it seemed like such an obvious answer that I wondered if you were just trolling. It appears you are in earnest, so here's my attempted answer:

When a flash crash such as this occurs, it is simply not the case that positions will be force closed before there is a possibility of lenders losing money. Say there are $16M in margin loans/open positions - when price dropped to $100 most of those positions would have passed their liquidation price, and would need to be liquidated. But the bid side has already been wiped out and so most of those positions are already deep underwater - liquidating into that order book would recover less USD than the loaned amount.

There is no way around this when leveraged positions greatly outweigh the bid depth and there is a sharp movement downwards - this is why the liquidity from Bitstamp is needed - so that the bid depth is greater. The 16M in loans, perhaps accounting for some ~25,000BTC (this assumes an avg entry price of 640, pretty conservative!) would have been backed by between 16M and 32M in collateral (assuming leverage is between 1 and 2). But some of that collateral was in BTC, not USD! Meaning that there is a double-whammy as BTC price drops - your margin balance decreases while your losses simultaneously go up!

So when price dropped from 630ish to 100, which looks to have been caused by the liquidation of around 7000 coins, for maybe $1.5 - $2M. Even if we assume that all the collateral was in USD, lenders for whoever got liquidated in that spike were already likely to be out of the money (7000 * 640 = $4.5M in loans, collateral of 4.5/2 = $2.25M plus liquidated value of $1.5-$2M. 2.25 + 2 < 4.5). Now if some of the collateral was BTC not USD, then the losses from that spike down to 100 look even worse.

But that was just the FIRST liquidation. The price drop would have triggered liquidations on most of the rest of the book. There were still 10M+ in loans, ie 15,000+ BTC, to be liquidated! With price at 100 or less, we get 15,000 * $100 = $1.5M, plus at most $5M = $6.5M, to pay back $10M in loans! And again I'm being super generous, assuming no BTC as collateral, whereas Raphael stated that a lot of users had BTC as collateral (for me it was about 50/50 when price was about 600). This makes the snap-back even worse, and could mean as little as perhaps $2-3M in recovered collateral leaving 3.5-5M only to pay back $10M+ in loans.

I have used best-guesses for a lot of these numbers, but based them on what I know of the amount of liquidity swaps, the price and the order book at the time of the flash crash so they should be ballpark. The parts I don;t know are the exact bid depth before the spike, the exact leverage ratio, and the ratio of USD to BTC collateral. But I think my analysis makes it pretty clear that there is plenty of risk for lenders to lose their money - I hope you can understand this now. Feel free to PM if you want more explanation.

Finally, it's worth mentioning that last night played out like it did because the BTC balance on Stamp ran out, meaning no more sells could be routed via Stamp. This is already a good reason to halt and wind back some of what happened. But circuit breakers, Stamp or no Stamp, may be worth having and it's good the community is discussing them.