In 2004 a Korean Airlines Cargo 747-400F in Oslo, Norway was loaded incorrectly and began its takeoff roll with a CG of 37.8% MAC. Thus being loaded 4.8% past the aft limit of 33% MAC.
The aircraft began to autorotate at 120 (KCAS) and thanks to the commander immediately suspecting a load error, the appropriate action was taken. However, nose attitude reached 19 degrees before sufficient downwards stab trim hab been applied.
B-HKD
... actually 58kts the nose wheel was off the ground. Aircraft was actually airborne at 120kts. On taxy to the runway the NLG WOG was intermittently air mode. The guys were the luckiest pilots on the planet.
FWIW, the Cm of the flaps is beneficial with an aft cg, MLG retraction is slightly beneficial. burning off the CWT may be wonderful for WBM structural considerations but guarantees the cg shifts aft. The arrival of that aircraft resulted in it departing the edge of the runway, however that is open to interpretation as the NLG was about 3' in the air, over the grass, but the MLG was on the concrete.
Had the AP disconnected in the cruise flight the aircraft probably would have been lost, as even with the AP engaged it exhibited longitudinal instability, and the elevator dP was rapidly cycling as was the elevator TE position in smooth air.
Not the first time, won't be the last time. Loading systems have many opportunities for variation from the expected process due to human ingenuity. A control problem close to the ground is a critical and generally untrained event. The opportunities for crosschecking are limited and need vigilance. I have bene caught out on the same type where 6.5T of cargo bound for the aft cargo comp went into the fwd comp, and we got to see the end of the runway up close and personal, took an extra 3000' of runway to get a rotate in, ended up with a part flap landing and full manual stab trip and still out of trim. Nowhere near as dangerous as the opposite case which appears to be a likely condition at Bagram.