Great thread! An interesting topic, one I've thought about a lot.
It seems to be basically related to peak cylinder pressure
Specific output imo does not really seem to be an accurate metric. "Specific output" refers to horsepower at one specific RPM (could be any RPM) but the only specific output that is ever quoted in the industry is peak HP, which occurs only at one RPM value a 'specific' RPM.
Let's use a Honda F20C as an example- it has a very high specific output for it's displacement and the engine only achieves that power output between (dependent on merket tune) 8300-8600rpm. It redlines at 9000rpm.
What happens if you mention to any casual the standard mantra "F20c produces 240hp from only 2.0L and they can generally last as long as a regular 2.0L engine in daily driving!" The casual's jaw will hit the floor like "Wow! Kool! Dude! What a feat So durable".
The average consumer has no concept of how engine powerbands work. Even a few seasoned veterans tend to forget.
The critical part of the story that is always missing is that the engine only actually produces 237-248bhp of output @+8500rpm. Yes, that's right, the quoted specific output has about a
500rpm window of duration in it's entire 9000rpm rev range. No one seems to remember that a standard 2.0L engine redlining at say 6500rpm will make more torque, and thus HP than an F20C from for it's entire 0-6500rpm range.
OTOH 4.3L Vortec peaks torque at about 2500rpm and makes about the same PEAK specific output as the F20C. The Vortec can produce that specific output much earlier, for a longer time and over a much broader RPM window. I strongly suspect the 4.3L would also have a lower BSFC/higher thermal efficiency at that output level also. If both engines spent their daily lives making within 20hp of their peak output, the F20C would not be dependable or economical and this is even considering that the Honda is a better made engine (metallurgy, clearances, balance, finishes, materials).
So generally IMO specific output is not a strong correlation to durability.
If anything, peak torque/unit of displacement would be a much better predictor because it correlates to peak cylinder pressures.
Therefore, in the case of engine A with a low peak torque against a high multiplier (RPM) sees much lower peak cylinder pressures per power stroke than engine B with a high peak torque against a low RPM multiplier (ultimately resulting in similar peak specific output), engine B will last a lot longer time outputting say a consistent 220hp. The difference is literally a 4.3L engine loafing at 3500rpm vs a 2.0L engine clinging to 8600rpm +/- 500rpm.