I think the combination of the torque modelling and tail wheel modelling, fubared some planes.
I've tried this: I found a way to kill the undercarriage but keep the engine alive. Do a slow speed, smooth rocking until the undercarriages fail and you will get a Spit9 with no gears, lying flat on the ground.
Now, in that state, apply full throttle and see what happens: the Spit9 starts to spin left like a helicopter. At full WEP, no amount of rudder will stop the spinning. The torque, is simply much larger than the rudder can handle.
What seems to happen when you try a take off in the P-51D or the Spit9, is the rudder input sort of messes with the plane's torque flow.
This torque, which pushes the plane into a tendency to yaw and roll left when stick left to idle(which also, in a plane with props rotating clock-wise, causes the plane to lift its right wings and wheels first, with the left wing and wheels lifting up later), with player input of the rudders, works against the torque flow and causes a massive spiking of the roll axis(since rudders themselves, also cause certain amount of movement of the plane in the roll axis).
This leads to a classic case of super-sensitive over-correction. The oscillation just increases massively, and the tail wheel modelling makes it worse. Because the plane is stable only when both the main gears are touching the ground, input of the rudders in Beta5 will cause one gear to momentarily leave the ground, causing the plane to swerve one way.
When rudder is released, reduced in input, or left to idle, over-correction kicks in and the plane rolls(rocks) to the other side, which causes the other wheel to leave ground, and the plane swerving in the opposite direction. This is because there's something fishy about how torque effects are modelled with the planes, no doubt about it.
If somehow players manage to gain enough speed before the main gear fails, and the tail wheel leaves the ground, only then the plane becomes stable. The methods suggested above, such as slow application of throttle or application of flaps, in essence, all rely on chance that the plane will be nursed to gain enough speed to let the tail wheel leave the ground, before the combination of weird torque, over-correction in the roll axis, and wild tail gears throw the plane over.
To make it short, the player has to hope that when taking off in the Spit9, he gets the correct combination of the micro-management of roll axis(to keep the main gears pressed to the ground until the tail wheel lifts) and yaw axis(so the plane stays at the runway).
If any of the two factors goes wrong, the plane fails to leave the ground. The problem is that correction of one axis, messes up with the other seriously. Due to the strange torque behavior, rudder input during take off, will cause the roll axis to become fubar, making one main gear come off the ground, and the other main gear+tail wheel causes the plane to swerve. Aileron input, coupled with the torque modelling, will cause over-correction and also make the plane swerve.
That means, none, of the above suggested methods will be consistent in giving satisfactory results for take off, with only the people with tremendous micro-management skills succeeding in take-offs with the same method over and over again. I've succeeded in taking the Spit9 off the ground with various methods, but never the same way twice - when one way works, if you try it again, it may fail.
Now, consider this: the torque, is obviously much larger in the Bf109G-10 than the Spit9. After take off, it takes nearly 3~4 seconds to roll right 90 degrees at 250mph, with a full 360 rightside roll coming near 10 seconds with just aileron input. If you drag the G-10 to vertical, when the speed falls under 200 mph right aileron input almost doesn't work at all. The plane refuses to roll right at all, even with full stick deflection.
Then, why is it easier to take off in the G-10 than the Spit9?
That's because the torque is so ghastly stronger than on the Spit9, that there isn't any oscillation or over-correction issues at all - the plane just mashes itself to the left, so there's no over-correction at all, no matter how you input the rudder or the stick.
....
That said, I don't think this can be right.
Some of the torque issues and tail-wheel modelling do add more realistic effects. That is true. Overall, more sensitive and difficult take-off/landing procedures are always welcome, and the chance of accidents being considerably high, makes even landings and take-offs an enjoyable experience for the gamer.
However, the torque factor and tail wheel factor, working in such a bizzare combination that a plane noted for ease of control such as the Spit9, almost fails to take off at all, definately means something's not right, or incomplete
....
As a bonus, I've found a way to take off in the Spit9, with 100% success rates.
1) After turning engine on, press alt+X and engage auto speed
2) Apply full throttle
3) When the speed nears 100mph, press X to engage level plane
4) No rocking, no yawing. The plane simply rolls forward straight. At 100 mph, the tail wheel is off the ground - at this point, you can take manual control and just pull on the stick and take off.
Watch carefully how the auto-trimming uses the trim. The rudder trim sways left and right continuously, which means the auto control is applying the precise amount of right rudder, to keep the plane going straight down the runway against the torque, and the two main gears staying solid against the ground at the same time. Unless you are really skilled, or have a very precise rudder system, you can't duplicate how the auto controls the rudder amount. It's literally a teensey amount of continuous rudder input that changes little by little, while the overall right rudder input is held at the same time.
