I love to hear your guess so please go ahead.
Totally a guess, and I am making many assumptions that you didn't say anything about -- hopefully they're in your logbook of tests.
Disclaimer: I'm a software engineer, not an EE. Point and laugh at my sophomoric assumptions.
You said that the smoke happened during your dummy load tests, IIRC. Also, IIRC, your dummy load is 33 ohms.
I am guessing your plot of the Magni 3, driven to distortion (1.5 W) was made at that time. The analyzer was doing a series of frequency sweep at that time? In any event, I'm guessing again that the Magni 3 was under load.
Playing around with Joule's Law and Ohm's Law (another assumption I'm making since with ac you can't assume this, but your dummy load should be purely resistive) we know that P=E^2/R. P=1.5, R=33. E^2 is therefore 49.5 (and E is approximately 7 V, but we don't care).
My big assumption is that you unplugged the dummy load under these conditions. A TRS (headphone) plug momentarily shorts a channel when you unplug the connector. The Magni 3 has no overcurrent protection.
You measured the output impedance of the Magni 3 at 0.6 ohms? If the Magni 3 was was delivering into a dead short, that would be it's "load." So P=E^2/R, E^2=49.5, R=0.6, the amplifier tried to deliver an impulse of 82.5 W over its traces. The caps don't care what the amp is rated at, and they're reasonably large.
I'm guessing that for a brief millisecond, the output transistor, traces on the board, and any resistors in that circuit were subjected to that momentary spike. Instant overheat. Instant smoke? Possibly instant solder splatter.
That makes a lot of assumptions that one cannot make. Some about what you did, some about how the circuit behaved, some DC formulae used in an ac circuit where they might not apply.
If one can crudely make that kind of guess without being inaccurate, the amp was damaged, and any results taken after the smoke came out are unreliable data.