I assume that the maximum reduction of 10% results from the haircut rule, which defines that the debt ratio may not exceed 10%. If the debt ratio were to be increased with a hard fork, a larger reduction would be possible. But what would happen to the haircut threshold then?

I agree. The haircut rule creates a 10% buffer above the current_supply that can be filled with SBD debt, and that's the only space where virtual_supply is eligible for reduction.

I guess the behavior after such a hardfork would also depend on other parameters, too. If it were just linear scaling like we have now, then I guess it would be more or less the same, just with the new cap. I haven't kept up with Hive, but I have the impression that they made some additional changes besides just increasing the max debt ratio. If we were to implement more complicated changes like that, I think it would probably need a whole new/different analysis.

However, a sample calculation showed that even here, with the price unchanged (above the haircut threshold), the virtual_supply does not change. This is also immediately apparent, as the changes to the STEEM supply and the SBD supply each depend on the current (unchanged) price.

True. I agree. It lowers the haircut price, but if the actual price is unchanged, then the lower haircut price doesn't matter for calculating the virtual_supply. By lowering the 80% threshold, it would move the reduction percentage to the right on the curve, which just means that the virtual_supply would be less sensitive to future price changes.

I just thought of an example of your second case (lower SBD supply):
If we did a hard fork to take the SBD in the DAO out of the debt calculation, it would have the same effect you showed.

Yeah, that's a far more realistic example than the one I used. It might actually be useful to think about implementing that. I guess it would knock the haircut price down by more than 50%, and it's probably justifiable, since those SBDs are all locked up, indefinitely.