but this can't be right.

Correct! That was, of course, a mathematically daring thesis. Of course, I can't draw conclusions about the fundamentals from one case. :-)
Indirectly, the goal can already be achieved. If the amount of the conversion is large enough so that the minimum price falls below the current feed reported by the witnesses, the haircut price no longer has an effect and the debt ratio actually decreases. This confirms your "update".

But now I've asked myself whether it's possible to find out when that will be?
The formula for calculating the minimum price after conversion is:

• sbd = current_sbd_supply before conversion,
• steem = current_supply before conversion,
• con = converted amount (in SBD),
• price = conversion price
• price_min = minimum price after conversion

Converted to con, which is the converted amount:
_{I just didn't have the desire to convert this formula on my own, so I asked ChatGPT to do it - and only checked it.}

This allows us to calculate how much SBD needs to be converted in order to undercut the current STEEM price and thus actually reduce the debt ratio:

Let's take the current price of 0.17524 $/STEEM (which isn't the internal feed price) and the current global properties:

or for typing:

con = ( 14,066,801.115 - ( 0.17524 / 9 × 483,387,043.866 ) ) / ( 0.17524 / ( 9 × 0.2619 ) + 1 ) = 4,332,613.037 SBD

So if someone currently converts more than 4.33 million SBD, this reduces the minimum price to such an amount that the debt ratio falls below 10% and SBD are printed again.
Understandably, the necessary SBD amount changes inversely to the STEEM price.