I'm not familiar with the circuit of the NT2 but from the sound of it, it seems likely that it is a faulty capacitor. The fact that it recovers after a few seconds of no audio makes it sound like the DC biasing is going AWOL with large signals. The favourites would be any electrolytic cap's which can go leaky or dry out. This could possibly be in the supplies derived from the phantom power.
The only other possibility I can think of, off the top of my head would be the mic insert biasing resistor (several gigohms) or the oscillator. I doubt this though as they should not really be affected by the audio level - but stranger things have happened.