The effective feed discount (and therefore the difference between actual and reported debt load is not as large as you suggest). This is due to the spread between feeds reported by various witnesses and how these are used to adjust the price, which in turn are not always adjusted in real time (rather than applying a median filter on the discount itself). At the time of my commenting now, the last reported (median) feed on the blockchain is 0.188 with a quote of 1.05. Comparing this with the current market price of 0.184, this implies a discount of only about 3%. This number varies over time but recently it has often been in the range of about 5-6%, due to several of the top 19 witnesses having reduced their discount (some using an automated process, in contradiction to the advice from the developer).

In any case, we can agree that the debt ratio is somewhere in the neighborhood of 5% (even 4% would be). As such accepting that as situation that doesn't call for reducing debt is playing Russian roulette (while slowly over time) squeezing the trigger six times, because this debt level will, especially if (incorrectly) not viewed as dangerous, be reached periodically and sooner or later, it is a certainty that a 50% price drop will occur turning 5% debt into 10% debt, haircuts, reputational damage to the entire Steem platform, and likely greatly increased overall instability that is difficult to characterize. Given a random walk model, we can expect a 50% price decline as equally likely to a 100% price increase. There is a 50% chance that the former, very dangerous, outcome will happen before the former, fortuitous and helpful outcome. By accepting this risk and betting on it not happening, we are essentially risking the heath of the system on a coin flip. (In fact given black swan-ish risks such as hacks of both the coin itself or exchanges or other large wallets, price pumps in BTC or some other coin that draws capital away, regulatory action, etc., large declines may in fact be more likely than large increases, but we can ignore that as the exact probabilities don't matter.) We got a bit of a reprieve the last time we approached 10% because a supportive community member who happened to be a very large SBD holder decided to convert a lot of it and power up. That community member has much less SBD now and may or may not want to power up even more. We can't always rely upon the goodwill of SBD holders to help rescue us, especially now that the amount of SBD in such friendly hands is now lower.

It may soon move down, though... your most recent price feed update would register a debt ratio of 3.06% ... or 2.71% if you were not discounting.

I will certainly be happy if the strong market action continues, the price increases and the debt ratio dwindles. However, we can't count on that. We can (and should) hope for the best, while preparing for the worst.

But, setting that aside... I have a much more substantive problem with how you conceive of the discount and its effects. I am very concerned with what seems to be your embrace of a permanent discount and your movement to reduce the interest on SBD

This is a mischaracterization of my position. I'm not in favor of a permanent discount (at least not one anywhere near as large as now, I do think think that a small, probably negligible, permanent discount may be necessary in some hypothetical distant future equilibrium but I think we can disregard that).

I am in favor of setting the discount based on the characteristics of a market where SBD hoarding is not being uselessly (and even harmfully) subsidized. In order to do that, the interest rate must be reduced first (or possibly in tandem) in order to assess how much discount is needed to maintain the peg when hoarding is not (or less) subsidized. The 10% initial rate made sense when the supply of SBD was very low and its value and supply needed bootstrapping (it would have been harmful to encourage conversions more than hoarding when the debt ratio was 0.01%). It is no longer the case that such a high rate adds value to the system when SBD is (we would I think at a minimum agree) at capacity and even the interest on it is itself increasing the debt.

I do not agree with the theory about the feed discount being directly tied to the "market expecting a decline". It is tied to some extent to the risk in the market (volatility) but if the market truly expected a decline, specifically, then people would sell now, at higher prices ahead of the expected decline, and the STEEM price would drop now, instead of a week from now. People able to do this consistently would make a large profit even in a generally declining market. Do you know of any? I don't. While it is theoretically possible that SBD trading market participants have a radically different forward-looking view of the STEEM price than STEEM market participants, it is pretty contrived and unlikely. This is especially the case because these are not really markets driven by "crowds". The number of participants are small and include mostly Steem users dumping their coins, a few somewhat sophisticated traders, and bots (sometimes these bots are indeed dumb, but I doubt that dumbness extends to making big bets on "expecting a decline" seven days in the future, something that would very likely fail to make a consistent profit both in backtesting and in live trading).

It is incorrect that the feed discount prices only the anticipation of a price decline. In fact, the discount prices a large (and not even fully knowable) variety of factors affecting or potentially affecting the willingness of market participants to engage in conversions including volatility/risk, cost of capital, availability of capital, liquidity and trading costs, overhead, possibility of changes to the feed discount itself, risk of exchanges shutting down wallet transfers, risk of exchange hacks or similar losses, quirks in the method of calculation (median of medians) of the conversion rate, quirks in feed reporting practices, risk of taking a 'haircut' (low now, but not zero), and probably others not sufficiently obvious to list.

These factors are difficult, if not impossible, to measure or assign a value ahead of time because they represent the hidden preferences of market participants and prospective market participants. That is why the white paper suggests increasing the feed discount when the rate of conversions becomes low (and debt is "dangerously high", a situation I consider to apply when SBD is only a little more than a 50% price drop away from taking a haircut, a process in turn fraught with many risks and costs to the system, which by the way is something you have also written about).

The witnesses can not directly control the actions of market participants. We can only provide incentives (in this case a feed discount) when needed. If the incentive is insufficient, then it should be increased; if it is sufficient, it should be maintained; if it is overly sufficient, it should be decreased. Price fixing that incentive based on what we think it is should be, and not based on what market participants actually do, is not correct or useful.

I wouldn't be so casual about suggesting SBD should be eliminated

I an absolutely not in favor of eliminating SBD either, but the possibility of viewing the experiment as having been tried and failed (as with hyperinflation) is something that people within the Steem community are increasingly beginning to discuss. Like you, I see SBD has having great benefit which is why I want to see its risks reduced and see it's supply being actively distributed to active platform users (including new ones) in the form of rewards, not hoarded by a relatively few well-off (and generally sophisticated) account holders. The former adds value by: 1) being more user-friendly to users recruited from outside of existing crypto circles who are uncomfortable with highly volatile and unfamiliar crypto assets (which is a top-priority goal and purpose of Steem); and 2) keeping it circulating and seeds a market for games, services, tips, gifts, commerce, etc. that will encourage its use. The latter does not.

In the discussion with the devs over whether witness rewards should be paid in SBD (since they are supposed to be used to pay for servers), it was explained that this was not a good idea, because the capacity to create SBD is limited and it is better to reserve that limited capacity for new users (where it provides the added value described above). This likewise applies to SBD that is hoarded, and also to the fact that SBD printing for rewards is stopped or reduced (instead giving STEEM to users which does not provide the added value described above).

I know you are in favor of removing or changing the 10% haircut rule, and depending on the situation after the fork I may also support that. However, given the system as it exists today, the 10% haircut rule exists and creates its own risks, risks which inherently make 5% debt ratio dangerously high. The same can be said for the print rate reduction that pays users in STEEM instead, and in doing so impairs the mission-critical goal of being a platform friendly to non-crypto users (and likewise if it turns out this appears safe to modify post-fork, I would support doing this as well). Dan has also, by the way, stated that he thinks that the current debt is quite high and that a 1% target is about right (which would restore SBD reward payments, although he didn't explicitly say that was his reason). If you look back at the how the SBD stability features such as reducing the print rate and then the haircut were described, it is clear they were intended as a remedial measure for high debt, not an intended operating point (though without action from witnesses this may unfortunately become the case). I disagree with him about plenty, but in this case I agree. As such, I can't support measures (high interest) which support the SBD price by incurring a significant cost to subsidize unnecessary and harmful hoarding and discourage debt reduction.