Keyholing is unexpected. Since bullets become more stable, and not less stable, as twist increases, and since other ammo shoots well in both guns, I suspect something is happening to these particular bullets in the faster twist. Your load level is middling unless you are using a magnum primer, in which case they could be nearer Hodgdon's maximum in actual pressure and velocity, depending on the primer brand, though that's still not quite up to the military's maximum.
My guess is you may be experiencing bullet core stripping. This is where angular acceleration by the rifling twist causes the jacket to spin up so quickly it breaks contact with the core, which then lags behind the jacket spin rate. When the bullet exits, the two rates of rotation find a compromise angular velocity as the the jacket loses spin to the core, and the core gains spin from the jacket. Being less massive than than the core, the jacket looses a lot more spin than the core gains in this exchange. The end result is a bullet spinning more slowly than the velocity and barrel twist would suggest it should be. The core is then also loose in the jacket, so the net projectile is poorly balanced and wobbles in flight and would tend to shotgun even if it weren't outright unstable.
The reasons I suggest core stripping are two-fold. One, you have pull-down bullets. The pulling process usually squeezes on the jacket some, and that squeezes the core. Copper gilding metal is more elastic than a lead core, so it springs back better. Add to this that the pulling was done to bullets already deformed by crimping, and you can see how it might weaken the contact friction between the core and jacket.
The second reason is that the bullets shoot well enough in a slower twist. All else being the same, the only thing the faster twist does differently is supply greater angular acceleration which increases the core stripping stress. If I'm right that this is core stripping, then I'd expect, unless they were made surplus because they couldn't pass accuracy testing in a 7: twist military rifle, that if the bullets were new and not deformed or pulled, they'd hold together for you.
Anyway, the core stripping hypothesis is easy to test. If you fire them in the 7" twist gun loaded for 7/8 as much velocity, they should then get the same rotational acceleration they do in the 8" twist gun. If they stop keyholing when you do that, the hypothesis is correct. Try 21.5 grains of powder and 19.7 grains of powder. I estimated by two different methods and got those two different numbers. Shoot the 21.5 grain load first, because if you find keyholing stops anywhere on the way down from below 24.5 grains, you have shown the bullets were likely to have been stripping.