There is a study of high speed bombers from the 1990's or early 2000's. There is a longer version but I did not find it. This is the short version. The thing to keep in mind here is that the entire system needs to be optimized, not 1 parameter. RAND looked at building a supersonic bomber not to be more survivable in the air, but to make the air bases more survivable by keeping them in CONUS.
1. The bombers still need RCS reduction to be survivable (obvious).
2. The bombers cannot be low IR emissions. Not a problem then, a huge problem now.
3. The bombers deliver low cost shorter range munitions.
4. The bombers need F-22 and tanker support.
5. The bombers are more efficient than subsonic bombers because they complete more sorties per unit time.
6. The study seems to ignore the entire history of supersonic flight, ie that whenever an air force or an airline has a choice, they choose subsonic. It can be said that USAF moved back to subsonic to increase bomber survivability (low altitude) but the airlines never went supersonic. In the end, the bomber is a delivery system. A study that trades off R&D cost, production cost, operating cost, munitions delivery per unit time, survivability, and the types of munitions that can be delivered due to that level of survivability has, to my knowledge, not been released. No RAND or CNA study that I have seen looks at this issue. It is germane to the issue we are discussing here though. A further complication is the fact that capabilities can be offloaded to other platforms much more easily than when the RAND study appeared. I suspect that the complexity of this study is why NGAD has been paused. I suspect that a technological development occurred that caused the USAF to redo the study of the type described here, a study they must have performed. The results of this study will not be available to us although we will see the results in NGAD. A further complication is that NGAD has to have A2A capability, a capability ignored in the RAND study. I think the sheer difficulty of developing a model, then crunching numbers while factoring in tech changed over the life of NGAD that will need to be accommodated in the program makes this vastly more complex than anything we have ever seen before.
The study below is a good start to see why speed is a good idea though.
RAND Supersonic Bomber Study