Yipes!

Yes, the base correlation model is a concern. More generally, instruments like these really require a much more careful analysis of the chaining of balance sheets in the economy, analyses of reserves, and checking that any so-called debt instruments aren't really equities or covered-called (as zero-down subprime mortgages were back in the day).

How close are the dominos to each other and how tippy are they?

What's wrong with synthetic CDOs? 

In my most humble opinion, there is nothing inherently wrong with synthetic CDOs. As I have stated elsewhere around here, the idea of reducing risk via diversification and levering it up or down via tranching are sound and useful concepts. What is questionable (and quite possibly "wrong") is the application of the usual derivatives business model of marking to model along with the associated upfront P&L recognition. So, synthetic CDOs, whether single-tranche or not, are probably just fine if done in a traditional structured credit context, with the usual caveat that pure ratings agency arbitrage may not add much social value.

Simple, clear and unchanged over time structure and 125 well defined underlyings - what's better for trading?

Simple, clear and unchanged over time structure and 125 well defined underlyings - what's better for trading?

I would tend to agree. Although the selection of the 125 matters, as does the weighting of them. There is a risk of irrelevance in comparison with typical portfolios. But this is not fundamental.

Synthetic CDOs are perfect for relative trading. You may set various ratios between tranches and the index and hedge. I'm missing this product.