It would depend on the frequencies involved, don't you think? A 1 m cable qualifies as a lumped component only up to about 30 MHz. USB 1.1 @11 Mbps isn't likely to care, USB 2.0 high-speed @ 480 Mbps might, and USB 3.0 definitely will. (Found a lil' article on the challenges of high-speed USB 2.0 on board layout from back in the day.)
That said, 90 ohms vs. 75 or 110 isn't a huge amount of mismatch (that's like -20 dB of s11 tops), so I wouldn't expect any really bad reflections. The 100 ohm twisted pair that goes in ethernet cabling would be an even closer match. I do have to wonder why they went with 90 ohms for USB, that's quite low for TP.
In fact, characteristic impedance is entirely independent of length by definition. Resistance, of course, will depend upon length and conductor diameter as you would expect, and a manufacturer could cheap out in ways that would push certain devices over the edge. (Some bus-powered interfaces were notorious for doing it even with good cables, like a certain Terratec that would develop a habit for crashing regularly... was it the 6fire USB? In this case proper operation would be restored by shorting out a choke in series with supply voltage, bypassing its resistance.)
Nothing spectacular would happen with USB 3.0 if you use 3m or 5m cable with so small impedance mismatch - you may only not be able to achieve top speed, but I doubt it. With audio this wouldn't present a problem, even with HiRes.
And I repeat once again, I believe 75 ohm and 110 ohm impedance were mentioned in the context of coax SPDIF, not USB.
I'm not sure what people expect to hear once impedance really starts to become a problem but the only sound they would hear in that situation would be the sound of silence.