I have connected a 1N4007 to the megger in reverse polarity and did a ramp test to 2.5 kV. Attached is the voltage/current graph (Current is microamps).
As can be seen, at about 1500V, the first degradation of the junction occurs. Right around 1800V, another punch through occurs. Soon after that, at 1850V, the megger declares breakdown because the current hits 3 mA, the megger's limit.
After this test, I have connected the same diode again to the megger and started the ramp again. The megger declared breakdown right away, at 100V which is the ramp start voltage.
No surprises so far. But here is what surprised me: I have measured the diode after the breakdown using two multimeters, in the "diode check" range, and the diode is measured to be in working condition with BOTH multimeters. I have also measured the diode in the ohms range, and it had a forward resistance of 38 KΩ (a good 1N4007 has a 40 KΩ forward resistance) and a reverse resistance beyond the multimeter limit.
So, the diode has basically failed in state where it MEASURES to be working, but fails when under voltage. And indeed, I applied 30V to it in reverse polarity, and it drew about 1 mA.
This means, when we measure high voltage diodes, just a multimeter measurement won't be conclusive. The diode needs to be tested with rated voltage applied to it