In general, I suggest that you search for Collin Moshman's videos about knockouts (direct links would be against the 2+2 rules) - he explains math adjustments in KO tourneys very well.
If there are $20 regs who use the chip EV mode Nash equilibrium in spots where the bounty player has gone all-in, they'll become $10 regs soon
In this spot, I think the GTO isoshoving range for SB is as wide as 70-80%, down to rags like 74s, because the other big stack has already folded and so the potential side pot doesn't deter SB from contesting the main one with the bounty added, and surely he'd love to remain in a HU pot with BTN.
As the bounty is worth 1/3 as much as the regular prize pool, 667 chips (2000 chips in play, divided by 3) should be added mentally to the main pot (whenever the bounty player is in and Hero is covering him) to represent the bounty. So it becomes t420 + t667 = t1087 here. The side pot is t380.
However, if Hero folds, then he doesn't totally lose the chance to take the bounty - if BTN wins the current hand, then, in the few next hands, Hero's chance of winning the bounty will be even slightly higher than 1/4 imo because the big stacks will make each other fold from time to time when the bounty player shoves again. I'd say it will be 30% or so.
BTN will win the current hand in 55-60% of cases if Hero folds because BTN's shoving range is stronger than chip EV Nash (consists of value hands only; the bounty player is inclined to wait for a better spot to get his chips in with a big equity and has an orbit or two to do so, as if he were surrounded by loose fish), I'd say it's top 20-25%.
Hence Hero's chance of taking the bounty later if he folds now is something like 60% * 30% = 18%, and we should add that to Hero's stack if he folds, i.e. assume that he'll have (t330 - t30) + 18% * t667 ~ t420 if he folds now.
If Hero calls all-in now and fails to win the main pot, his chance of winning the bounty later becomes negligible in any case.
So the problem is reduced to whether to call all-in to contest the main pot of t1087 and the side pot of t380 vs a top 25% (BTN) and a top 70% (SB) range or to fold and retain t420.
In this model, QTo becomes a snap call, and actually, Hero should call with a 50%-ish range imo, something like this: