"These amplifiers go to eleven...."
The answer is complicated. It depends on gain, noise, overload and resolution. Your ears are your best guide when things get complicated.
For digital to digital, I leave everything at max, and let my preamp deal with volume. The Mytek DAC I use has an analog volume control I only use for the headphones.
However, say the Oppo players. You can set the volume of any speaker from several DB above, to below unity (input = output). The problem is, and they acknowledge it, setting the speaker levels above unity causes digital distortion because you need more bits than the players have. This is also a problem for upsampling players. If you upsample a 24 bit master, you may end up creating samples louder than 24 bits will allow. Benchmark media has a blog about this.
Digital volume controls can be from crappy to decent. The problem is they all involve a lot of math, so you loose the bit-perfection of the original signal as soon as you apply any sort of multiplication or division. This is a big problem with computer sound, and why so much discussion is around avoiding the normal audio mixers built into the OS.
Some of the best digital volume controls however use 32 bit DAC’s to compensate for the loss of resolution you would have otherwise.
In the analog domain, your issues are things like channel matching, overload and noise. Any analog output has a limited power supply, which defines the maximum voltage swing. Most preamps also have a lot more gain available than you'll ever need. The more gain, the more noise. Turn the volume up and you'll run out of voltage, causing clipping. Poor volume controls cause imbalance in L to R matching, and the wipers they use can get dirty.
