I'm not being sarcastic -- but how do you know that paralleling the
outputs will produce a significant increase in power? Ignoring the
very pertinent question of how to do this /correctly/, the most you
could get is a 3dB increase, which is plainly audible, but hardly
a large increase.

A small increase is better than nothing.

No, it isn't. Not when it requires a lot of work, and will be only
slightly
audible.

Fitting resistors is hardly a lot of work. I just need someone who knows
their
stuff, to answer my original question.

So, William, what is the correct answer here (value of resistors)? I'm
curious now, too.

I don't see where I'm obliged to provide an answer, but here's what I
believe, to the best of my understanding...

Assuming that both channels have the same output impedance, and the output
levels are closely matched, then (if the outputs are directly paralleled)
there will be /no/ current flow from one channel into the other, and the
total amount of current that /could/ be delivered will be doubled. But if
the load impedance remains the same, there will be no increase in current
flow and no increase in volume level. You would need a speaker of half the
impedance to gain more power.

I think that's correct. Anybody want to agree or disagree?