Marc,
I regard bandwidth less than, say, 75kHz as highly suspect. Not that I claim to be able to hear that high as a discrete tone, but there have been numerous studies showing that it's audible on things with a sudden rise time (drums, cymbals, what have you--no guarantees on flutes). Straight Fourier analysis of a complex waveform shows harmonics going *way* out. Turns out it's audible (hell, even the numbers folks can get behind this--go look up the formula for a square wave).
Keep in mind that 75kHz is less than two octaves beyond 20kHz. Most circuits are good down to 1 or 2Hz on the bottom...a good 3 to 4 octaves below 20Hz.
A somewhat more subtle problem is that where there's a rolloff, there's a phase shift. And the phase shift is present (and measurable, for those who care to do so) way below the -3dB point. This is why you see equipment with specs for phase shift at 20kHz; sometimes also at 20Hz. Phase shifts are often overlooked because you've got to have a scope to see them. But if you bring in a waveform (let's assume lagging, in the case of high frequencies) an eighth of a cycle behind the rest of its companions, you're going to get transient phase-related frequency cancellations. I.e. your frequency response is no longer flat, and the waveform you put in ain't what comes out, hence a transient distortion mechanism that's completely ignored by steady state tests. Square waves are a good start for testing such things; impulse testing is better still.
The long and the short of it...I'll take as wide a bandwidth as I can get. (With a nod towards stability, let's go for the 100-200kHz range.)
Dialectric absorbtion and dissipation factor, too. Charge an electrolytic to rated voltage. Let it sit for a minute. Discharge it quickly, using a low value resistor. Then hook a meter to it and watch the voltage as it climbs. Voltage? *What* voltage? There's no voltage there...you just discharged it, right? It was at 0V. Only, it isn't. It's at 1.5V and still climbing. Now 2V. Those electrons were hiding in the dialectric. Now they come creeping back out. Now 2.5V. This is not good. We have a passive component acting as an active component, injecting current flow into the signal that shouldn't be there.
Try it. If nothing else, put a 9V battery on a 15V cap. (Although it's easier to see at higher voltages.)
Then try the same thing with a good film cap.
Have I heard good sounding equipment with electrolytics in it? Yes. And progress has been made on electrolytics, true. But like I said, the best case scenario for electrolytics I've ever heard were a few times where it was only a slight improvement, hardly worth the (admittedly very high) price for good film caps. Lower price film caps may or may not be such an improvement, though I've got a small box of old Sprague Orange Drops that I toss into things when I'm short of the good stuff. Even humble polyester has been an improvement in most cases. (I eventually order in something decent and toss the Orange Drops back in the box until the next time.)
Grey