The modern paradigm of mathematical finance is that the One True Way to manage risk is to create risk indices, create liquid markets for assets based on those indices, and then use those assets for hedging risk exposures.

CDS fit the bill as long as they never trigger but the fiction that they might trigger is maintained by all concerned. Then, the "CDS spread" is a risk measure with an associated more-or-less-liquid asset, the CDS, which people can use to hedge the sensitivity of their pricing models to credit risk.

As soon as CDS start triggering, selling CDS becomes a risky proposition and the market for CDS becomes a lot less liquid, making it unsuitable for hedging and potentially leading CDS spreads to decouple from any "fundamentals-based estimate" of credit risk.

If credit events start happening and CDS don't trigger, the sale of CDS is exposed as a scam in the eyes of the genuine credit risk hedgers (those concerned about default losses, not about daily noise in prices and CDS spreads) and not only the market for CDS becomes illiquid again, but also people suddenly realise that <gasp> credit risk is not hedgeable, and in the current banking risk paradigm credit-risky business lines are abandoned.

Recall "Anglo-Saxon" banking abhors credit risk. The investment banking model prefers transaction-based or underwriting businesses where small amounts of capital are turned over at high speed in high-margin transactions. The ongoing collapse of Project Finance due to the problems of the Continental European banking system illustrates this: "Continentals" do credit risk (even in "Anglo-Saxon" markets), or used to until the current crisis hit.

There are three stories about the euro crisis: the Republican story, the German story, and the truth. -- Paul Krugman