Re: "In my case, I move my RPM up specifically to make the trans survive. If I tuned for a 3200rpm HP peak, I can trash the trans. If I move it to 4400rpm it lives, even though I make more HP by tuning up there."
Thanks, that's something I was aware of - but didn't know it applied to LSR.
The problem seems to be that (absent really high flywheel inertia) the input shaft and all whirly bits never "see" continuous flow of power, especially with low inertia parts - they get a series of really big hammer blows to the engagement surfaces (splines, flutes, dogs, blah). At lower speed, and with smaller number of cylinders the compression cycling hits it in the opposite direction.
This is true even with torque converters (although I suspect the pulse amplitude decays to near zero with stall speed) - the crank flange "ratcheting" behavior is still presen at the input shaft despite no mechanical connection.
If you can't reduce the amplitude (shock value) the next best thing is increase the frequency, and frequently moving the max power upscale does both (-10% torque but +20% RPM = more power and better gear life).
It also appears that a higher engagement number helps to some degree. Assuming equal execution, A/T drums with 54 splines take more abuse than 36 spline, dual or triple row small pitch chains last longer than equal nominal tensile load large pitch.
If the vehicle is traction limited, the maximum load the transmission sees won't be as high as if the tires didn't slip, but:
1. the engine still puts out all the transient power used to accel the bits right down to the tire
2. breaking traction > backing off > opening throttle cautiously > breaking traction > lather, rinse, repeat introduces load reversals to all the engagement points - always a bad thing.
3. IMHO the Tor-Sen stuff proves that helical gears don't like being loaded backwards, which is what happens to all diffs (Ford 9" especially bad) when the car is "dragged down" by the tires after breaking traction.