Hello, could someone please advise me how to solve a cross conduction/thermal runaway problem?

I have designed a power amp to deliver around 20W into a 1 Ohm load at 200KHz sine. The amplitude is more or less constant, the frequency is constant, the load varies a bit. The supply rails are +/-12V. My design uses 2 or 3 pairs of output transistors (2 pairs did not work well, tried 3 pairs still no go).

The amplifier works perfectly until we reach a point where the load gets too heavy (too small) in conjunction with a high amplitude signal. At that moment, we get immedaite thermal runaway so fast that the heat hardly makes it out of the transistor cases and cross-conduction - chicken and egg - I am not sure what causes what.

The problem is immediately fixable by not letting the load drop so low, or by simpoly adding another volt to the supply rails. I attach 3 scope snapshots taken at the emitter resistors.

The first picture shows normal operation, load is around 2-2.5 Ohms. You can see the transistors opening and closing properly without overlaps.

Second picture shows what happens when the load gets too low, a few seconds in the transistors have already remained open for far longer in the cycle than they should resulting in thermal runaway and current limit kicking in.

Third picture shows the problem corrected by increasing the rails by 1.5V.

If anyone has experience with RF power amps please help.

Thanks