The losses incurred by holders of real estate-backed loans are real. These loans were handed out at the peak of the bubble to people who could not afford them at unsustainable rates. When low downpayments met a collapse in housing prices, many borrowers faced loan values greater than their home values and walked away. This effectively made billions of dollars invested in the commercial paper that was a key workhorse in international finance worthless.
Before the crisis, this fact was evident to many people. Several hedge funds attempted to short the market for such loans, and were destroyed. I know of at least one person, however, who managed to time his trades right and made quite a few billion dollars. However, most people did not face the responsibility that those traders did and continued to buy and sell worthless paper. It's not the "models" that are at fault; other models that actually corrected for the unique credit qualities of recent borrowers would give you higher predicted defaults. Rather, the originate and distribute model reduced the incentives to model correctly and allowed mortgage originators to hand out loans to poor borrowers and resell them into a global savings glut.
So again, the losses from handing out such loans is high. But it's perhaps on the order of $500 billion--severe, but similar to the asset collapse from any bubble. The problems from this crisis, however, have spread into many parts of the finance industry and into the real economy as well. It's a big question how a relatively manageable loss nearly broke down the entire financial system.
One explanation I gave earlier is that marking assets to market overstated the problems for certain firms and caused a crisis in confidence. That's a part of the answer to why there was a spiral, though such transparency is good in general. Here at my day job, many people are interested in this question and I came across other ways to get from "manageable loss" to "end of the world".
A key factor is the extraordinary extent to which modern finance relies on debt. Many people imagine that you can show up on Wall Street and easily make money if you have flexible ethics. Generally though, most people face very efficient markets and thus need to ramp up on leverage to amplify any small inefficiency they find. Investment Banks were among the worst offenders, and their debt was very short-term to boot. As banks faced continual deterioration in their balance sheets, financial requirements forced them to raise more capital. An interesting solution to this problem is Rajan's proposal to replace required capital deposits with catastrophe insurance for banks. The idea is that forcing banks to keep a certain amount of liquidity on the books prevents that capital from being used precisely at the hour of need, so instead banks could contract out for funding conditional on facing a crisis.
There are other liquidity spirals that amplify original shocks. The complexity of the financial toxins spread around induced significant asymmetric information and constraints on knowledge. People don't know what's going on and can't find out, so they reject entire asset classes as worthless in bad times. Finally you end up with classic bank runs when people lose faith in financial institutions altogether. This happened not just in the case of Northern Rock and IndyMac, but was a factor in Bear Sterns, Lehman Brothers, and the rest of the Investment Banks. Tightly linked financial markets run on trust between counterparties, so you have giant network effects that blow up any perceived risk. See James Surowiecki on this point.
The financial system was so highly leveraged and tied together that blaming subprime mortgages is a little besides the point. The entire financial system was in something of a bubble, and was one asset collapse away from destruction. Much smarter regulation will be required to fix and prevent future crises, and that's unlikely to materialize.