Just hypothesising here. Your squirrel-cage rotor is made of laminated steel, with the aluminium/copper ladder (the squirrel cage) embedded in it. For it to work as an induction motor the laminated steel on the rotor should not itself be magnetised - the force comes by the induction of massive currents in the squirrel cage (aluminium/copper) by the stator, which causes a magnetic field interaction between rotor and stator and causes the torque. The currents in the squirrel cage are caused by the "slip" between the rotating magnetic field created by the stator and the rotor speed. So normally when you run it at full power, with 50 Hz applied, it spins at slightly less than 50*60/2=1500 rpm, i.e. 1400 rpm, which creates a 100*2/60 = 3.3 Hz electrical "slip" frequency and the currents in the rotor are 3.3 Hz (and huge).
But, if your rotor happened to get permanently magnetised somehow, then it would turn into a hybrid between an induction machine and a synchronous machine. In a synchronous machine, the torque is not proportional to "slip" but proportional to the angle between stator magnetic field and rotor position. In this case, useful torque is only created when the slip is zero and the rotor moves exactly in synchronism with the stator magnetic field. It's just possible that if this has happened to your rotor, then as you try to drive it fast, it starts slipping (due to friction) and the useful torque drops to zero. I've never heard of this happening, but perhaps the rotor is accidentally (permanently) magnetised somehow.
Its common in synchronous machine to include squirrel cage windings to aid in starting, and to introduce damping to reduce rotor oscillations. But, I never heard of anyone deliberately introducing a permanent-magnet (synchronous) machine behaviour into an induction machine.
This might not be whats going on in your machine, but its a remote possibility.
Andrew.