There are two chief issues in play here, I think:

1. I believe CIYAM's point is that there is nothing backing the stabilisation of the currency, except for the currency itself, which is the same way bitshares works. If demand crashes, there could be a downward spiral of devaluation causing a black swan. This is largely a fair point, which can only be resolved by having the backing currency off chain, which is hard to say the least.

2. Having the blockchain as a market maker subjects the currency itself to the adverse selection problem due to 'informed traders', which all marker makers face. This scenario manifests itself as traders buying from you when you ask is too low and selling to you when you bid is too high. There are a proportion of informed traders present in any market, alongside 'noise traders' who buy/sell at any price. Your simulation won't have included the effect of informed traders because it only manifests itself in live forward testing. However, you can add it to your simulation - informed traders know with certainty the future direction of the price after time period X, so they profit from your pricing mistakes.

edit: I meant to write about the difference between point 2 and bitshares approach, which you might find helpful - bitshares doesn't have a blockchain market maker, instead it socialises the adverse selection risk, because individual users lock up collateral (equivalent of your buffers) in order to create their pegged assets, and other users are required to take the other side of the trade. This way, the currency itself doesn't inflate wildly out of control in the case of a black swan, instead individual users get margin called.