It looks like colored coins have the ability to completely replace actual bitcoins as units of value, and in fact this might be preferable. I am surprised I didn't think of this myself because the first (or maybe second or third) thing I thought of when hearing about bitcoins was that a currency based on a issuer-backed market-basket of goods would have been superior, but much more difficult to bootstrap (how do you build trust in a currency that anyone can issue on a network with no enforcement mechanism?). Having an existing infrastructure to pass new coins of arbitrary value sounds like a Win, and bootstrapping the first colored coin value against bitcoin value is smart.

Do you think it's possible that fiat $10,000/USDcoins2 colored coins redeemable for either $10,000 OR 100 bitcoins would be economically feasible? There would be no demand if there was an expected movement from $100/bitcoin since the issuer wins on volatility, but what if a distribution of distinct USDcoins2 were issued, with 1% being redeemable for $15,000 or 150 and the remaining 99% redeemable for $10,000 or 100 bitcoins, but whose actual value would only be revealed after a period of time (months). Pre-committing to this seems fairly straightforward and would generate demand for the currency with a positive expected net value. I believe the effect would be toward stabilizing the exchange rates at ~$100/bitcoin and 100 bitcoins/USDcoins2. I don't know if the exact percentages would work, but I imagine that there is some distribution that would have the intended effect. The issuer would be spending to achieve a stable exchange rate, which may be important for clearinghouses, banks, or individuals with savings.

EDIT: I forgot to clarify that the issuer of USDcoins2 decides whether to pay $10,000 or 100 bitcoins when an owner redeems 1 USDcoins2.