What is the explanation of non-zero VCE voltage of an IGBT while its gate is high?

Question

While investigating a malfunctioning high power induction heater system (which constantly destroys its IGBT modules), I came across an interesting situation.

Circuit

Gate drivers are SKYPER 32 PRO R and IGBT modules are SKM300GB12T4. There are snubber capacitors on every IGBT's Collector-Emitter terminals. Resonance frequency is ~6.4kHz. The driver module has a built-in overshoot protection and it's enabled. Minimum dead time is set to 1µs.

During the investigation we found out that the short circuit protection function of the gate driver was permanently disabled by connecting the Vc_sense pin to Emitter. So the gate driver can only see zero volts between Vc_sense and Ve.

I thought that enabling this function could prevent the IGBT destructions. However, as we enabled the short circuit detection function by making the required wiring, system (the gate drivers) immediately switches to fault state above 4.5V of DC bus bar voltage.

We disabled the protection again and run the system with a low power, current limiting power supply. While system is correctly functioning, we measured the V_GE and V_CE voltages with an isolated ground scope. Blue is V_GE and red is V_CE of A1 IGBT:

At this moment, the heater heats the sample iron block, indicating the system is working as it should.

Question

As can be seen, when gate is high, V_CE stays ~0V for 10µs and then rises like a sinus wave. How can this happen? Shouldn't V_CE stay at ~0V while its gate is high?

Further experiment

I suspected that it's because of the L//C load and the user side can not drive the gate drivers in a correct frequency/phase, so the load might behave like a power source. In order to eliminate this possibility, I disconnected the L//C load and connected a group of incandescent light bulbs as a pure resistive load.

Results were somewhat similar:

The Voltage source is 55V and current is 130mA during this light bulb experiment.

Answer 1

As I'm collecting data upon request of bunker89320, I suspected some other signals. I didn't expect to see any signal between Driver.A1.Emitter and A1.Emitter (the EC point) but when I measured, I saw a weird, half sinus like signal with a 100V peak voltage. When I double checked the application, I realized that the physical locations of IGBT's and my schema are not matched.

Apparently we'd been keeping connecting A2.C to Driver.A1.V_c_sense pin and so on. After fixing this, the signals VGE and VCE observed as expected:

Lessons Learned

Cables weren't numbered and the components weren't labeled in the panel. It was obvious at the first glance that there was a serious documentation problem of the system (any documentation problem is a serious documentation problem). I should have ignored the time pressure of the customer and strictly required completing the documentation before starting any kind of debugging session.