I ordered the new IGBT's from Mouser, after managing to poke around with my meter probes without disassembling anything, and found one was a dead short.
Once the spares arrived, i took out the old ones and discovered that interesting milling(?) pattern in that marking window, which "got my spider senses tingling"...
Before finding that failed-open resistor on channel A, i had a temporary issue on channel B, which turned out to be a wonky connection for the feedback taken from the output connector.
The weird thing is, at that point, i had no oscillation visible on channel B, despite a larger output signal level.
And yes, the zobel resistors both look fine, and measure as they should (10 ohms).
Good, that you replaced the devices with the correct parts. If the old ones were fakes, they wouldn't have lasted 5 minutes in operation under load.
It's a matter of narrowing it down I suppose.
It could be the channel just repaired is oscillating, and affecting the other channel, or power supply noise common to both channels, or a fault in the line level input processing circuitry common to both channels.
I haven't had a look at the manual yet, but I assume the output zoebel networks are intact?