3/ One key detail is the buyback does not come out of sUSDe yield alone and it is not funded from a separate pot. It becomes another line in the fee allocation, sized as a share of gross revenue and split alongside the existing lines rather than ahead of them. So every allocation line shrinks by the same percentage. A 10% take means an even take from sUSDe stakers, partner payouts and Aave liquid leverage. Those three lines are where the revenue goes today, at roughly a third each.

4/ Result one, the size. Had the design been live over the past two years, it would have been switched on for 118 days out of 705. USDe spent the rest of the time below the first rung. During the months it was running, it would have bought $52.7M of ENA a year. Averaged across the whole two years, idle stretches included, $8.82M a year. For scale, ENA gross unlocks run at about $512M a year. So the buyback offsets roughly a tenth of that while running on historical data. The proposal converts each rung into dollars using an illustrative 6% protocol yield. The actual yield over the past year was 4.6%, so we ran all our figures at the realized rate for a comparison.

5/ Result two, what it costs sUSDe holders. On the 93 active days where sUSDe was paying more than sUSDS, it still paid more than sUSDS after the buyback on 82 of them. On the other 11 it slipped just behind, all of them within a tenth of the benchmark. Small and worth understanding why it happens. Here is why it happens at all. The buyback takes its 10% whenever USDe is large enough. It never checks whether sUSDe can afford to give up 10%. Usually it can, because sUSDe pays well above sUSDS and losing a tenth still leaves it ahead. On the days sUSDe is only just ahead, the same 10% puts it behind. For context, sUSDe already paid less than sUSDS on 243 of the 705 days we looked at, with no fee switch running at all. This is a market condition the design lives with and not the condition it creates.